Risk, Uncertainty and Investment Decision-

Making in the Upstream Oil and Gas Industry

Fiona Macmillan
MA ons !University o" A#erdeen$
Octo#er %&&&
A thesis presented "or the degree o" 'h(D( at the
University o" A#erdeen
D)*+ARA,IO-
This thesis, and the research underpinning it, is entirely my own work. It has not been
submitted in any previous application for a degree. All quotations in the thesis have
been distinguished by quotation marks, and the sources of information specifically
acknowledged.
Signed:..
i
A*.-O/+)DG)M)-,0
!hilst this thesis is entirely my own work, many others have contributed to it and
shaped the end result in their own unique way and I would like to take this
opportunity to recognise them.
"irst, thanks must go to my sponsor, #SI$% Australia &#ommonwealth Scientific and
Industrial $esearch %rganisation', and supervisors for their enthusiasm, patience and
commitment, especially (rofessor )raeme Simpson, whose confidence in me and in
the value of my study never wavered even when I doubted it myself. Second, I would
like to acknowledge the respondents and the companies that they represented. "or
their time, honesty and encouragement I am e*tremely grateful. Special thanks must
go to Steven +c#oll of #onoco, ,on )luyas of -asmo, (at +ackintosh of ./0 &.et
/orske 0eritas' and )illian .oyle of !ood +acken1ie for all their assistance. -ast
and most importantly, I would like to thank my husband, +ike, my parents and my
friends 2 especially )ill, 0anessa and /atalie, for their love, support and unshakeable
belief in me and in all that I do. !ithout them, there would be no thesis.
ii
A10,RA*,
The research presented in this thesis is rooted within the e*isting decision theory and
oil industry literatures. It contributes to one of the current debates in these literatures
by providing evidence that in the operators in the 3.4. upstream oil and gas industry
there is a link between the use of decision analysis in investment appraisal decision2
making by organisations and good business performance.
It is commonly acknowledged that decision analysis is not as widely used by
organisations as was predicted at its conception &for e*ample, Schuyler, 5667'. %ne
reason for this is that no study to date has shown that use of decision analysis
techniques and concepts can actually help individuals or organisations to fulfil their
ob8ectives. .espite over four decades of research undertaken developing decision
analysis tools, understanding the behavioural and psychological aspects of decision2
making, and applying decision analysis in practice, no research has been able to show
conclusively what works and what does not &#lemen, 5666'.
The current study begins to fill this gap by using qualitative methods to establish the
following. "irstly, the research identifies which decision analysis techniques are
applicable for investment decision2making in the oil industry, and thereby produces a
description of current capability. Secondly, the study ascertains which decision
analysis tools oil and gas companies actually choose to use for investment appraisal,
and through this develops a model of current practice of capital investment decision2
making. -astly, using statistical analysis, it provides evidence that there is an
association between the use of decision analysis in investment decision2making by
companies and good organisational performance in the upstream oil and gas industry.
Such research not only contributes to the current theoretical debate in the oil industry
and decision theory literatures but also provides valuable insights to practitioners.
iii
*O-,)-,0
'AG)
Declaration i
Ackno2ledgements ii
A#stract iii
+ist o" "igures vii
+ist o" ta#les viii
*hapter 34 Introduction 3
5.5 Introduction 9
5.9 :ackground to the thesis 9
5.; $esearch questions <
5.< %utline of thesis =
*hapter %4 +iterature Revie2 3&
9.5 Introduction 55
9.9 $isk and uncertainty 55
9.; #urrent practice in investment appraisal decision2making 5=
9.< The evolution of decision theory 9>
9.? .ecision analysis and organisational performance ;5
9.@ #onclusion ;7
*hapter 54 ,he Oil Industry in the U(.( 56
;.5 Introduction <>
;.9 #urrent challenges in the global oil industry <>
;.; The oil industry in the 3.4. <7
;.< Investment appraisal decision2making in the oil industry ?9
;.? #onclusion ?<
iv
*hapter 74 Methodology 88
<.5 Introduction ?@
<.9 Adopting an appropriate methodological framework ?7
<.; Avaluating the effectiveness of the research methodology @6
<.< #onclusion 75
*hapter 84 *urrent capa#ility in investment appraisal in the upstream oil
and gas industry 9%
?.5 Introduction 7;
?.9 The concepts of e*pected monetary value and decision tree analysis 7<
?.; (reference theory =@
?.< $isk analysis 6=
?.? (ortfolio theory 5>?
?.@ %ption theory 555
?.7 #urrent capability 55=
?.= #onclusion 59?
*hapter :4 *urrent practice in investment appraisal in the upstream oil
and gas industry 3%:
@.5 Introduction 597
@.9 The use of decision analysis by organisations 59=
@.; The investment appraisal decision2making process 5<>
@.< A model of current practice 5?9
@.? #onclusion 5??
*hapter 94 ,he relationship #et2een the use o" decision analysis in investment
appraisal decision-making and #usiness success4 a non-parametric
analysis 389
7.5 Introduction 5?=
v
7.9 The type of study 5?6
7.; $anking companies by use of decision analysis tools and concepts 5@5
7.< $anking companies by organisational performance 5@6
7.? (roposing the hypotheses and selecting the statistical tests 57?
7.@ $esults 577
7.7 .iscussion 576
7.= #onclusion 5=>
*hapter ;4 *onclusion4 #et2een <e=tinction #y instinct> and <paralysis #y
analysis> 3;%
=.5 Introduction 5=;
=.9 The research questions revisited 5=;
=.; Theoretical contribution 5=6
=.< Implications of the study to practitioners 56;
=.? "uture research 56<
=.@ #onclusion 56@
Appendi= 34 Intervie2 0chedule 366
Appendi= %4 'resentations and 'apers %&:
Appendi= 54 ,he 0pearman *orrelation ,est %&;
Appendi= 74 ,he .ruskal /allis and /ilco=on Rank 0um tests %&6
Appendi= 84 *ritical values o" "or 0pearman tests %37
Appendi= :4 *ritical ?alues o" . "or .ruskal /allis test 2ith 5
independent samples %38
Appendi= 94 *ritical ?alues o" *hi-0@uare at the &(&8 and
&(&3 level o" signi"icance %3:
Appendi= ;4 *ritical ?alues o" 0 "or the /ilco=on Rank 0um ,est %39
1i#liography %3;
vi
+I0, OF FIGUR)0
'AG)
;.5 !orldwide giant fields <9
;.9 #ampbellBs prediction of world oil production after 566@ <;
;.; .istribution of remaining &Cet2to2(roduce' oil &in :illions of :bls'
by country <<
;.< .istribution of remaining &Cet2to2(roduce' oil &in :illions of :bls'
by region <<
;.@ Actual spot :rent oil price over time <@
;.7 The average si1e of 3.4. fields by discovery year <=
;.= .iscoveries by field2si1e class in the /orth Sea <6
;.6 !orldwide operating costs ?>
?.5 The upstream oil and gas industry: a multi2stage decision process
7?
?.9 #umulative cash position curve 7@
?.; Typical spider diagram 76
?.< An e*ample of a decision tree
=9
?.? A preference curve
==
?.@ Typical preference curves
=6
?.7 Analysis using A+0
69
?.= Test results eliminated 6;
?.6 The decision2makerBs preference curve 6;
?.5> Analysis using preferences
6<
?.55 $educing risk through diversification
5>@
?.59 A 6-step approach to investment appraisal in the upstream oil
and gas industry 556
vii
?.5; #hoke model
59<
@.5 A model of current practice
5??
7.5 A decision tree 5@@
=.5 The relationship between decision analysis and behavioural decision
theory 569
=.9 :est practices in organisationsB use of decision analysis 56;
+I0, OF ,A1+)0
'AG)
9.5 #onceptualisations of risk and uncertainty 59
?.5 .iscounted cash flow concept 7=
?.9 Dypothetical field data 66
?.; Dypothetical field data for +onte #arlo simulation 66
?.< $esults from the +onte #arlo simulation 5>>
?.? :ase value data and probability distribution assigned to each of the
reservoir parameters 5><
?.@ Table of the output generated using the base value data and input
distributions specified in table ?.?. 5><
?.7 Safe and risky pro8ects 5>6
?.= All possible outcomes of investing ?>E in each pro8ect 55>
?.6 The similarities between a stock call option and undeveloped reserves 55?
@.5 %rganisationsB use of decision analysis 5?>
7.5 $anking of companies by their use of decision analysis techniques
and concepts 5@=
7.9 $anking of companies by performance criteria 57<
7.; Spearman correlation coefficients between performance variables
and use of decision analysis 577
viii
*hapter One
Introduction
5
3(3 I-,RODU*,IO-
The aim of this chapter is to introduce the research pro8ect and to outline the research
themes that guide the study. The research presented in this thesis is rooted within the
e*isting decision theory and oil industry literatures. It contributes to one of the
current debates in these literatures by providing evidence that in the operators in the
3.4. upstream oil and gas industry there is a link between the use of decision analysis
in investment appraisal decision2making by organisations and good business
performance.
3(% 1A*.GROU-D ,O ,) ,)0I0
$esearch into decision2making has become increasingly popular over the last forty
years, and many published studies now e*ist &for e*ample, "ord and )ioia, 9>>>F
)unn, 9>>>F Akenberg, 9>>>F +ilne and #hanF 5666F /utt, 5666, 5667 and 566;F
:urke and +iller, 5666F (apadakis, 566=F .ean and Sharfman, 566@F Guinn, 56=>F
+int1berg et al., 567@F #yert and +arch, 56@;'. !hilst, these studies are useful for
providing broad insights into the field of decision2making, very few have investigated
investment decision2making in comple* business environments where there is
substantial risk and uncertainty and each investment decision requires significant
capital e*penditure without the prospect of revenues for many years.
.ecision analysis &$aiffa, 56@=F Doward, 56@=F $aiffa and Schlaifer, 56@5' is a label
given to a normative, a*iomatic approach to investment decision2making under
conditions of risk and uncertainty &)oodwin and !right, 5665'. :y using any one, or
a combination, of decision analysis techniques, the decision2maker is provided with
an indication of what their investment decision ought to be, based on logical argument
&#lemen, 5666'. (revious research into the usage of decision analysis by companies
has typically been survey2based and produced evidence of a difference between the
decision analysis techniques described in the literature, and the decision analysis tools
which practitioners choose to use &for e*ample see studies by Arnold and
Dat1opoulous, 5666F #arr and Tomkins, 566=F Schuyler, 5667F :uckley et al., 566@
"letcher and .romgoole, 566@F Shao and Shao, 566;F 4im et al., 56=<F Stanley and
:lock, 56=;F !icks 4elly and (hilippatos, 56=9F :avishi, 56=5F %blak and Delm,
9
56=>F Stonehill and /athanson, 56@='. It appears that whilst decision analysts
describe a range of decision analysis techniques, some of which are very
sophisticated, organisational decision2makers are choosing to utilise only the most
simplistic tools and concepts in their investment decision2making &Atrill, 9>>>'.
Dowever, the methodological approaches adopted by the researchers conducting these
studies precluded them from providing any e*planation into the reasons why some
techniques fail to be implemented and others succeed &#lemen, 5666'. #onsequently,
some writers, typically behavioural decision theorists such as Tocher &567@ and 567=
reprinted in "rench, 56=6', have e*plained the results by arguing that decision2makers
choose not to use decision analysis techniques because their use adds no value to
organisationsB investment decision2making processes since decision analysis does not
aim to predict what decision2makers will do, only to suggest what they ought to do.
#lemen &5666' offers another interpretation. De believes that at least one reason why
decision analysis techniques and concepts are not widely used by organisations is that
no study to date has provided evidence that organisations that use decision analysis
tools perform better than those companies that do not. .espite over four decades of
research undertaken developing decision analysis tools, understanding the behavioural
and psychological aspects of investment decision2making and applying decision
analysis to practical e*amples, no research has been able to show conclusively what
works and what does not. #lemen &5666' believes that to rectify this situation, future
studies into investment decision2making should investigate the relationship between
organisational performance and the use of decision analysis techniques. If, as many
decision analysts believe &for e*ample, "rench, 56=6', companies that use decision
analysis in investment decision2making outperform those that do not, such research
would contribute to the theoretical debate between the decision analysts and
behaviouralists. The behavioural decision theorists would no longer be able to claim
that there is no value in a theory that does not aim to predict what decision2makers
will do. Such research would obviously also be valuable to practitioners.
This type of study, however, has been slow to appear in the literature doubtless
because of the threat they represent to the decision analysts &#lemen, 5666 pp9;29<':
HAsking whether decision analysis works is risky. !hat if the answer is
negativeI The contribution will clearly be scientifically valuable, but many
;
individuals J consultants, academics, instructors J with a vested interest in
decision analysis could lose standing clients, or even 8obs.K
The current study aims to remedy this situation by researching the use of decision
analysis in investment appraisal decision2making by the ma8or companies in the
upstream oil and gas industry. The oil and gas industry epitomises investment
decision2making under conditions of risk and uncertainty &!atson, 566=F /ewendorp,
566@F $ose, 56=7F Ikoku, 56=<', and hence was one of the first industries to apply
decision analysis &)rayson, 56@>'. The industry is often used as a laboratory for the
development of new decision analysis tools and concepts &for e*ample, :ailey et al.,
in pressF )alli et al., 5666F :all and Savage, 5666F .i*it and (indyck, 566= and 566<F
Smith and +c#ardle, 5667' and it is recognised to lead all other industries, with the
e*ception of the finance industry, in the e*tent to which it uses decision analysis
&Schuyler, 5667'. #learly, then the oil industry provides a particularly useful conte*t
in which to establish whether a relationship e*ists between the use of decision
analysis in investment appraisal by companies and business success. The study will
focus on those ma8or upstream oil and gas companies that are operators in the 3.4..
Since most of the ma8or oil companies that operate in the 3.4. are global players in
the oil industry, the findings will be indicative of investment decision2making in the
worldsB ma8or upstream oil and gas companies. The research questions that the thesis
aims to answer and methodological approach followed are outlined in the following
section.
3(5 R)0)AR* AU)0,IO-0
3( /hich techni@ues are the most appropriate "or companies to utilise in their
investment decision-makingB
This question is motivated by the observation that there are many decision analysis
techniques presented in the academic investment decision2making literature leading
many practitioners to feel confused about which decision analysis techniques are most
applicable for investment decisions &see #hapter @ and studies by Schuyler &5667' and
"letcher and .romgoole &566@''. #learly, there is a need to identify which of the
decision analysis techniques and concepts presented in the academic investment
<
decision2making literature are the most appropriate for practitioners to use for
investment decision2making. The current study undertakes such research in the
upstream oil and gas industry.
The current study draws on the decision analysis and oil industry literatures to
ascertain which decision analysis tools are the most appropriate for companies to use
for investment decision2making. This involves firstly, identifying the whole range of
techniques that are available and, secondly deciding which of these tools are the most
appropriate for upstream investment decision2making. This demands careful
consideration of factors such as the business environment of the upstream industry
and the level and type of information used for investment decision2making in the
industry. Through this process, the research identifies the decision analysis
techniques that are particularly useful for upstream investment decision2making. This
constitutes current capability. Then, drawing again on the investment appraisal and
industry literatures, and also on insights gained at conferences and seminars, an
approach to investment decision2making in the oil industry is presented that utilises
the full spectrum of tools identified. Some decision analysts advocate using one
decision analysis technique for investment appraisal &for e*ample, Dammond, 56@7'.
Dowever, in reality, each tool has limitations &-efley and +organ, 5666' some that are
inherent, others which are caused by a lack of information or specification in the
literature. As such, the knowledge that the decision2maker can gain from the output
of one tool is limited &/ewendorp, 566@'. Therefore, a combination of decision
analysis techniques and concepts should be used to allow the decision2maker to gain
ma*imum insight which, in turn, encourages more informed investment decision2
making. Some oil industry analysts have recognised this and presented the collection
of decision analysis tools that they believe constitute those that decision2makers ought
to use for investment decision2making in the oil and gas industry &for e*ample,
/ewendorp, 566@'. Dowever new techniques have only recently been applied to the
industry &for e*ample, )alli et al., 5666F .i*it and (indyck, 566= and 566<F $oss,
5667F Smith and +c#ardle, 5667' and as such, these previously presented approaches
now require modification.
%( /hich techni@ues do companies use to make investment decisions and ho2
are they used in the investment decision-making processB
?
This question is prompted by the observation highlighted in section 5.9 that very few
previous studies into decision2making have investigated the use of decision analysis in
investment appraisal decision2making by organisations. The current study e*amines
the use of decision analysis in investment appraisal decision2making within the
operating companies in the 3.4. upstream oil and gas industry.
.ata are collected by conducting semi2structured interviews in twenty2seven of the
thirty2one companies who were operators in the 3.4.Bs upstream oil and gas industry
in +arch 566=. The data is analysed in two stagesF first against the core themes
contained in the interview schedule &Appendi* 5', which are informed by the
literature analysed in #hapters 9 and ;, and the emergent themes identified in
contemporaneous notes taken during the research process. Second, after this initial
coding, the data is coded again. In this second level coding, the core themes are more
highly developed and closely specified, and other emergent themes are included. This
allows the researcher to develop a model of current practice in investment decision2
making in the upstream oil and gas industry that is grounded in the data. The model
provides insights into the use of decision analysis in investment appraisal decision2
making organisations. In particular it permits identification of the techniques
organisations do use and those that they do not, and, by drawing on the behavioural
decision theory literature and the interview data, it is possible to suggest reasons for
this.
5( Is there a relationship #et2een using decision analysis techni@ues in
investment appraisal decision-making and good organisational per"ormanceB
This question is motivated by the observation by #lemen &5666' discussed in section
5.9 that there is a need for researchers to e*plore the relationship between the use of
decision analysis in investment appraisal decision2making by companies and
organisational performance. The current study investigates whether such a
relationship e*ists in the operating companies in the 3.4. upstream oil and gas
industry.
@
0ery few other studies have attempted to value the usefulness to organisations of
using decision analysis &#lemen, 5666'. Some studies in behavioural decision theory
have evaluated the effectiveness of individual decision analysis techniques &for
e*ample, Aldag and (ower, 56=@F ,ohn et al., 56=;F Dumphreys and +c"adden,
56=>'. Dowever, such research has been criticised because the studies typically use
hypothetical decision situations and there is evidence in the behavioural decision
theory literature to suggest that people make different decisions under these
circumstances than the decisions they would make if the situation were real &Slovic,
566?F )rether and (lott, 5676F -ichenstein and Slovic, 5675F -indman, 5675'.
#lemen and 4wit &9>>>' investigated the e*istence of a relationship between use of
decision analysis and organisational performance in 4odak. The researchers used
depth interviews and documentary analysis to inform their research. This
methodological approach permitted the researchers to value the HsoftK effects on the
organisationBs performance of utilising decision analysis techniques and concepts.
Dowever, whilst their research provides useful insights, as the authors themselves
acknowledge, the focus on one organisation meant that the results could not be
generalised to a larger sample. The current study differs from this since it attempts to
establish whether there is a relationship in the operating companies in the 3.4. oil
industry between using decision analysis in investment decision2making and business
success. Therefore, by implication, the research involves numerous companies and
this prohibits use of the type of time2consuming qualitative methodology
implemented by #lemen and 4wit &9>>>'.
Instead, the current study uses the indication of current capability and current practice,
gained from answering the first and second research questions, to rank the operating
companies according to the number of decision analysis techniques they use for
investment appraisal. The research then assumes that any value added to the company
from using a decision analysis approach, including any HsoftK benefits, ultimately
affects the bottom2line. This means that it is therefore possible to use publicly
available financial measures and other criteria indicative of performance in the
upstream oil and gas industry, to indicate business success. The e*istence of a
relationship between organisational performance and use of decision analysis in
investment appraisal decision2making in the oil industry is then analysed statistically.
7
The remainder of the thesis concentrates on answering these research questions. Aach
chapter is outlined in the following section.
3(7 OU,+I-) OF ,)0I0
The literature review in #hapter 9 draws on the academic literature on investment
decision2making to highlight the gaps in the e*isting literature that the research
questions presented above are drawn from. It is structured so that attention is
focussed on the source of each of the research questions in turn.
#hapter ; draws on the oil industry literature to provide a brief description of the
conte*t of the current study that highlights the main challenges facing the oil industry
in the 95
st
century. Since the current study is located in the 3.4., the effects of these
global changes on the 3.4. oil industry are e*amined. This indicates the growing
comple*ity of the industryBs business environment and highlights why it is such a
useful environment in which to study the use of decision analysis in investment
decision2making.
#hapter < outlines the methodology adopted in the research. The current study
utilises qualitative methods for data collection and a combination of mechanisms for
data analysis. The qualitative method of semi2structure interviewing is used for the
investigation of companiesB investment decision2making processes and non2
parametric statistical analysis is employed to investigate the relationship between the
use of decision analysis in investment appraisal decision2making and organisational
performance. Aach type of analysis is evaluated in terms of their appropriateness for
the study of investment decision2making.
!hilst #hapter ? primarily draws on secondary data sources, it is presented as a
significant contribution to this thesis, since it first identifies the decision analysis
techniques available for upstream oil and gas industry investment decision2making,
and also presents a new approach to investment decision2making in the industry
which utilises this spectrum of tools.
=
#hapter @ presents the first set of findings from the research interviews. It draws on
the interview data to provide a model current practice in investment decision2making
in the upstream oil and gas industry. In particular, the decision analysis techniques
that upstream organisations actually use are presented. !hen this is compared with
the indication of current capability ascertained in #hapter ?, the findings confirm the
trend observed in previous quantitative research studies that there is a gap between
current theory in investment appraisal and current practice. Dowever, unlike these
survey2based studies, where the research methodology used prohibited further
investigation of such issues, the current study uses insights from the semi2structured
interviews, together with behavioural decision theory literature, to suggest why this
might be the case.
#hapter 7 uses the data presented in #hapters ? and @ to produce a ranking of the
companies according to their usage of decision analysis techniques in investment
appraisal decision2making. The assumption that any value added to the company
from using a decision analysis approach will ultimately affect the organisationBs
bottom2line is 8ustified. This assumption is then used to investigate the relationship
between the ranking of organisations by their use of decision analysis in investment
appraisal decision2making and business success statistically by using criteria that are
indicative of organisational performance.
The final chapter, #hapter =, brings together the information gathered for the thesis
and provides the answers to the research questions posed in #hapter 5. It sets out the
conclusions that can be drawn from the research. In particular, the implications of the
results to the theoretical debate between the decision analysts and behavioural
decision theorists are highlighted. The limitations of the research presented in this
thesis are discussed and this leads into the identification of areas for future research
that arise from the current study.
6
*hapter %
+iterature Revie2
5>
%(3 I-,RODU*,IO-
This chapter presents the literature review for the current study. It draws on the
e*isting academic literature on investment decision2making to highlight the gaps in
this literature that the research questions presented in #hapter 5 are drawn from. The
literature review is structured so that attention is focussed on the source of each of the
three research questions in turn.
%(% RI0. A-D U-*)R,AI-,C
The first section of the literature review emphasises the centrality of risk and
uncertainty to investment decision2making by focusing on the following three
questions:
5. Dow does the academic investment decision2making literature conceptualise risk
and uncertaintyI
9. Dow do investment decision2makers conceptualise risk and uncertaintyI
;. Dow do these decision2makers cope with risk and uncertainty in investment
decision2makingI
Investigating the methods of coping with risk and uncertainty adopted by investment
decision2makers highlights the role of quantitative techniques. This leads into
identification of the need for a study that ascertains which of the tools and techniques
that are presented in the decision theory literature are most appropriate for investment
appraisal. This is the first research question that this thesis aims to answer.
#onsider the first question proposed above. $isk and uncertainty are inherent in all
decision2making &:ailey et al., in press, Dammond et al., 5666F Darrison, 566?F
)oodwin and !right, 5665F +organ and Denrion, 566>' and hence receive
considerable attention in the academic investment decision2making literature &for
e*ample, Atrill, 9>>>F :uckley, 9>>>F +urtha, 5667F :orsch and +ossinF 56@='. This
prominence is well deserved. 3biquitous in realistic settings, risk and uncertainty
constitute a ma8or obstacle to effective capital investment decision2making &Simpson
et al., 9>>> and 5666F -amb et al., 5666F :all and Savage, 5666F !atsonF 566=F $ose,
55
56=7F +urtha, 5667F /ewendorp, 566@F %ransanu and #onnolly, 566;F +c#askey,
56=@F :runsson, 56=?F #orbin, 56=>F Thompson, 56@7'.
AU,OR0 ,)RM *O-*)',UA+I0A,IO-
5. Anderson et al. &56=5'
9. Anderson et al. &56=5'
;. Anderson et al. &56=5'
<. Anderson et al. &56=5'
?. Dumphreys and :erkley &56=?'
@. -athrop and !atson &56=9'
7. -athrop and !atson &56=9'
=. +ac#rimmon and !ehrung
&56=@'
6. Darrison &566?'
5>. Darrison &566?'
55. Spradlin &5667'
59. Dolmes &566='
5;. Dolmes &566='
3ncertainty
3ncertainty
$isk
$isk
3ncertainty
$isk
3ncertainty
3ncertainty
$isk
3ncertainty
$isk
$isk
3ncertainty
A situation in which one has no knowledge about
which of several states of nature has occurred or
will occur
A situation in which one knows only the
probability of which several possible states of
nature has occurred or will occur
Same as &5'
Same as &9'
The inability to assert with certainty one or more
of the following: &a' act2event sequencesF &b'
event2event sequencesF &c' value of consequencesF
&d' appropriate decision processF &e' future
preferences and actionsF &f' oneBs ability to affect
future events
(otential for deleterious consequences
-ack of information available concerning what the
impact of an event might be
A*posure to the chance of loss in a choice situation
A common state or condition in decision2making
characterised by the possession of incomplete
information regarding a probabilistic outcome.
An uncommon state of nature characterised by the
absence of any information related to a desired
outcome.
The possibility of an undesirable result
A situation which refers to a state where the
decision2maker has sufficient information to
determine the probability of each outcome
occurring.
A situation where the decision2maker can identify
each possible outcome, but does not have the
information necessary to determine the
probabilities of each of the possibilities.
Table 9.5: Conceptualisations of risk and uncertainty (source: adapted from Lipshitz and Strauss,
1997
59
Dowever, despite this prominence, there is much confusion in the academic
investment decision2making literature over the definitions of risk and uncertainty.
Table 9.5 presents a sample of the definitions of risk and uncertainty given by some of
the contributors to the capital investment decision2making literature. The table clearly
illustrates conceptual proliferation in the academic investment decision2making
literature. This has led Argote &56=9 p<9>' to assert:
Hthere are almost as many definitions of risk and uncertainty as there are
treatments of the sub8ect.K
A comment echoed by Cates and Stone &56=9 p5':
Hif we were to read 5> different articles or books on risk, we should not be
surprised to see it described in 5> different ways.K
To answer the first question proposed above of how the academic investment
decision2making literature conceptualises risk and uncertainty then, it is clear that
whilst it is widely acknowledged in this literature that risk and uncertainty are
inherent in capital investment decision2making, there is no conceptual basis for
agreement of the definitions of risk and uncertainty.
The second question that this section aims to address, how investment decision2
makers conceptualise risk and uncertainty, has received relatively little attention in the
empirical literature on investment decision2making &-ipshit1 and Strauss, 5667'.
Dowever, there is evidence in this literature which suggests that the conceptualisation
of risk and uncertainty adopted by a decision2maker affects the method of coping that
the decision2maker adopts &-ipshit1 and Strauss, 5667'. +illiken &56=7' found that
decision2makers encountering diverse risks and uncertainties respond differently. The
e*istence of contingent coping is a recurrent theme in the academic decision2making
literature &for e*ample, )ans, 5666'. #yert and +arch &56@; p556' proposed that:
HLorganisationsM achieve a reasonably manageable decision situation by
avoiding planning where plans depend on prediction of uncertain future events
and by emphasising planning where the plans can be made self confirming
through some control device.K
5;
)randori &56=<' specified which of five decision2making methods should be selected
given the magnitude of risk and uncertainty caused by a lack of information.
Thompson &56@7' specified which of four decision2making approaches should be
selected given the amount of risk and uncertainty. :utler &5665' later adapted this
model.
To answer the section question of how investment decision2makers conceptualise risk
and uncertainty then, the empirical investment decision2making literature offers many
hypotheses and scant empirical evidence regarding how decision2makers
conceptualise risk and uncertainty &-ipshit1 and Strauss, 5667'. Dowever, it does
indicate that the definitions of risk and uncertainty that are adopted by decision2
makers affect the model or mechanism they use to handle risk and uncertainty
&-ipshit1 and Strauss, 5667F :utler, 5665F )randori, 56=<F Thompson, 56@7'.
The last question that this section aims to address, how decision2makers cope with
risk and uncertainty, follows from this and has received considerable attention in the
investment decision2making literature &for e*ample, #lemen and 4wit, 9>>>F #lemen,
5666F )ans, 5666F )alli et al., 5666F -ipshit1 and Strauss, 5667F +urtha, 5667F
/ewendorp, 566@F $aiffa, 56@=F $aiffa and Schlaifer, 56@5'. According to Smithson
&56=6 p5?;' the prescription for coping with risk and uncertainty advocated in much
of the capital investment decision2making literature is:
H"irst, reduce ignorance as much as possible by gaining full information and
understandingSecondly attain as much control or predictability as possible
by learning and responding appropriately to the environment"inally,
wherever ignorance is irreducible, treat uncertainty statistically.K
Thompson &56@7' suggests that organisations constrain the variability of their internal
environments by instituting standard operating procedures and reduce the variability
of e*ternal environments by incorporating critical elements into the organisation &that
is, by acquisition or by negotiating long2term contractual arrangements'. Similarly,
Allaire and "irsitrotu &56=6' list several Hpower responsesK used by organisations to
cope with environmental uncertainty including shaping and controlling e*ternal
events, passing risk on to others and disciplinary competition. Dowever, the standard
procedure for coping with risk and uncertainty advocated in the investment decision2
5<
making literature is outlined in the section of this literature referred to as decision
theory &#lemen and 4wit, 9>>>F #lemen, 5666F )oodwin and !right, 5665F "rench,
56=6F $aiffa, 56@=F Doward, 56@=F $aiffa and Schlaifer, 56@5'.
In the decision theory literature, the process decision2makers are advised to adopt for
coping with risk and uncertainty involves three steps known as $.G.(. &-ipshit1 and
Strauss, 5667'. The first stage involves the decision2maker reducin! the risk and
uncertainty by, for e*ample, conducting a thorough information search &4aye, 566?F
.awes, 56==F ,anis and +ann, 5677F )albraith, 567;'. The decision2maker then
"uantifies the residue that cannot be reduced in the second step. "inally, the result is
plu!!ed into a formal scheme that incorporates risk and uncertainty as a factor in the
selection of a preferred course of action &/ewendorp, 566@F Smithson, 56=6F Dogarth,
56=7F #ohen et al., 56=?F $aiffa, 56@='. Aach step will now be discussed further.
This will highlight the role of quantitative techniques and introduce the concept of
decision analysis. The section will conclude by identifying the need for a study that
ascertains which of the many decision analysis tools and concepts described in the
decision theory literature are the most appropriate for investment decision2making.
This is the first research question that this thesis aims to address.
Strategies for reducing risk and uncertainty include collecting additional information
before making a decision &4aye, 566?F .awes, 56==F )albraith, 567;F ,anis and
+ann, 5677'F or deferring decisions until additional information becomes available
and it is possible to reduce risk and uncertainty by e*trapolating from the available
evidence &-ipshit1 and Strauss, 5667'. A typical method of e*trapolation is to use
statistical techniques to predict future states from information on present or past
events &:utler, 5665F Allaire and "irsirtou, 56=6F :ernstein and Silbert, 56=<F
!ildavski, 56==F Thompson, 56@7'. Another mechanism of e*trapolation is
assumption2based reasoning &-ipshit1 and Strauss, 5667'. "illing gaps in firm
knowledge by making assumptions that go beyond, while being constrained by, what
is more firmly known which are sub8ect to retraction when, and if, they conflict with
new evidence, or with lines of reasoning supported by other assumptions &#ohen,
56=6'. 3sing assumption2based reasoning, e*perienced decision2makers can act
quickly and efficiently within their domain of e*pertise with very little information
&-ipshit1 and :en Shaul, 5667'. Scenario planning, imagining possible future
5?
developments in script2like fashion &Schoemaker, 566?', is another strategy of
reducing risk and uncertainty that combines prediction and assumption2based
reasoning. "inally, risk and uncertainty can also be reduced by improving
predictability through shortening time hori1ons &preferring short2term to long2term
goals, and short2term feedback to long range planning, #yert and +arch, 56@;', by
selling risks to other parties &Dirst and Schweit1er, 566>', and by selecting one of the
many possible interpretations of equivocal information &!eick, 5676'.
It is important to recognise, however, that reducing risk and uncertainty by collecting
information can be problematic since often the information is ambiguous or
misleading to the point of being worthless &Dammond et al, 5666F +organ and
Denrion, 566>F "eldman and +arch, 56=5F )randori, 56=<F !ohstetter, 56@9'.
+oreover, there is evidence to suggest that collecting information does not help the
decision quality when the level of environmental uncertainty is very high &"redrickson
and +itchell, 56=<'. This leads some to adopt an entirely different approach to
reducing risk and uncertainty by controlling the sources of variability that decrease
predictability. "or e*ample, as discussed above, according to Allaire and "irsitrotu
&56=6', some organisations use Hpower responsesK &-ipshit1 and Strauss, 5667'.
It is the second stage of the $.G.(. heuristic that much of the decision theory literature
discusses &for e*ample, #lemen and 4wit, 9>>>F Dammond et al., 5666F #lemen,
5666F Thomas and Samson, 56=@F 4eeney, 5676F 4aufman and Thomas, 5677F $aiffa,
56@='. .ecision analysis &$aiffa, 56@=F Doward, 56@=F $aiffa and Schlaifer, 56@5' is
a normative discipline within decision theory consisting of various techniques and
concepts that provide a comprehensive way to evaluate and compare the degree of
risk and uncertainty associated with investment choices &/ewendorp, 566@'.
Traditional methods of analysing decision options involve only cash flow
considerations, such as computation of an average rate of return &/ewendorp, 566@'.
The new dimension that is added to the decision process with decision analysis is the
quantitative consideration of risk and uncertainty &#lemen and 4wit, 9>>>F #lemen,
5666F /ewendorp, 566@F )oodwin and !right, 5665F +organ and Denrion, 566>F
"rench, 56=6F $aiffa, 56@=F Doward, 56@=F $aiffa and Schlaifer, 56@5'. In #hapter ?,
all aspects of decision analysis will be discussed in detail and specific techniques will
be reviewed. Dowever, for the purposes of gaining an overview of the approach, the
5@
standard decision analysis can be summarised as a series of steps &Simpson et al.,
5666F -amb et al., 5666F /ewendorp, 566@F )oodwin and !right, 5665F +organ and
Denrion, 566>F "rench, 56=6F Thomas and Samson, 56=@':
5. .efine possible outcomes that could occur for each of the available decision
choices, or alternatives.
9. Avaluate the profit or loss &or any other measure of value or worth' for each
outcome.
;. .etermine or estimate the probability of occurrence of each possible outcome.
<. #ompute a weighted average profit &or measure of value' for each decision choice,
where weighting factors are the respective probabilities of occurrence of each
outcome. This weighted2average profit is called the e*pected value of the
decision alternative, and is often the comparative criterion used to accept or re8ect
the alternative. Another measure that can be used to compare decision alternatives
is the e*pected preferenceNutility value of the decision alternative. This is a
decision criterion that attempts to take into account the decision2makerBs attitudes
and feelings about money using preference or utility functions. In either case, the
decision rule is to choose the decision alternative with highest e*pected
preferenceNutility value. This is the third and final stage of the $.G.(. heuristic.
The new parts of this standard decision analysis approach are steps ; and <
&/ewendorp, 566@'. The analyst is required to associate specific probabilities to the
possible outcomes. Since this basic approach was proposed, the e*perience gained by
academics and consultants has stimulated changes designed to make the decision
analysis approach more fle*ible to the needs of managers &for e*ample, Dammond et
al., 5666F Thomas and Samson, 56=@F 4eeney, 5676F 4aufman and Thomas, 5677'.
$ecently, as computing power has increased, the dimension of simulation has been
added to the standard decision analysis approach &/ewendorp, 566@'. $isk analysis
based on +onte #arlo simulation is a method by which the risk and uncertainty
encompassing the main pro8ected variables in a decision problem are described using
probability distributions. $andomly sampling within the distributions many, perhaps
thousands, of times, it is possible to build up successive scenarios. The output of a
risk analysis is not a single value, but a probability distribution of all e*pected returns.
57
The prospective investor is then provided with a complete risk2return profile of the
pro8ect showing the possible outcomes that could result from the decision to stake
money on this investment &/ewendorp, 566@'.
+ore recently, preference, portfolio and option theories have been attracting some
attention in the decision theory literatures &for e*ample, :ailey et al., in pressF
Simpson et al., 9>>>F Simpson et al.O 5666F )alli et al., 5666F Dammond et al., 5666F
Smith and +c#ardle, 5667F $oss, 5667'. Aach of these techniques will be discussed
in #hapter ?. The plethora of techniques that are presented in the academic decision
theory literature for the quantification of risk and uncertainty has confused
practitioners &see Section @.; of #hapter @ and studies by Schuyler &5667' and
"letcher and .romgoole &566@''. +ost decision2makers report uncertainty about
what each tool aims to do, the differences between techniques and are unclear about
when certain tools should and should not be used &Section @.; of #hapter @'. #learly
then, there is a need to identify which of the decision analysis techniques and concepts
presented in the academic decision theory literature, are the most appropriate for
investment decision2making. The current study aims to do this by answering the first
research question which was posed in #hapter 5.
The focus in this chapter now turns to the motivation for the second research question
proposed in #hapter 5. In e*ploring this question the researcher aims to ascertain
which techniques companies actually use to quantify risk and uncertainty in
investment appraisal and to understand how the results from the techniques are
plugged into the organisational investment appraisal decision2making process. The
following section draws on the academic investment decision2making literature to
analyse the recent studies of current practice in investment decision2making. In doing
so, it identifies the gap in the e*isting literature that by answering the second research
question and producing a description of current practice in investment appraisal in the
operators in the 3.4. upstream oil and gas industry, this study aims to fill.
%(5 *URR)-, 'RA*,I*) I- I-?)0,M)-, A''RAI0A+ D)*I0IO--MA.I-G
The fundamental concepts used in decision analysis were formulated over two
hundred years ago. Cet the application of these concepts in the general business
5=
sector did not become apparent until the late 56?>s and early 56@>s &for e*ample,
)rayson, 56@>', and it has only been within the last five to ten years that it has
seriously been applied to investment decision2making in practice &for e*ample, see
Section @.; of #hapter @ and studies by Schuyler &5667' and "letcher and .romgoole
&566@''. "urthermore, it is widely acknowledged that current practice in the
techniques used for investment appraisal decision2making in practice in all industries
trails some way behind current decision theory &for e*ample, Atrill, 9>>>F Arnold and
Dat1opouous, 5666F Schuyler, 5667'. This has been established via empirical
research which has tended to focus on whether, when and which decision analysis
techniques are used by organisations &for e*ample see studies by Arnold and
Dat1opoulous, 5666F #arr and Tomkins, 566=F Schuyler, 5667F :uckley et al., 566@
"letcher and .romgoole, 566@F Shao and Shao, 566;F 4im et al., 56=<F Stanley and
:lock, 56=;F !icks 4elly and (hilippatos, 56=9F :avishi, 56=5F %blak and Delm,
56=>F Stonehill and /athanson, 56@='. These studies have typically used survey
techniques to produce statistical results indicating the percentage of organisations
using decision analysis techniques &for e*ample, Schuyler, 5667'. As will be
discussed in more detail in #hapter <, utilising survey techniques for data collection
has precluded the researchers from conducting an investigation of why companies
endorse the use of some techniques and yet fail to implement others and, more
importantly, it prevents the identification of the decision analysis techniques which
perform best &that is, where the predicted outcome from the technique is close to the
actual outcome' &#lemen, 5666'. As will be seen in section ;.<, the failure of these
earlier studies to investigate such issues has contributed to the divide between the
behavioural decision theorists and decision analysts, and to the gulf between current
practice and current capability in decision analysis highlighted above &#lemen, 5666'.
Avidently then, since the empirical research conducted to date has limitations, there is
a need for a study to establish common practice in investment appraisal. This is the
second research question that this thesis aims to address.
The current study will use a qualitative methodology. This will allow the researcher
not only to establish which decision analysis techniques companies are currently
using, but also to investigate other, HsofterK issues. "or e*ample, if the study confirms
the earlier empirical studies that there is difference between the techniques described
in the academic investment decision2making literature &which will be identified by
56
answering the first research question proposed in #hapter 5' and those which
companies choose to use, it will e*plore this issue. "urthermore, since previous
research has suggested that the relationship between the conceptualisation of risk and
uncertainty in the organisation and the techniques or method of coping with risk and
uncertainty adopted by decision2makers &see section 9.9', this will also be
investigated. The researcher will then be able to offer insights into how the results
from the decision analysis techniques are integrated into the organisational investment
decision2making process.
Attention is now focussed on the source of the third research question which aims to
establish whether there is a relationship between the use of decision analysis
techniques by organisations and organisational performance. The ne*t section
e*amines the evolution of the decision theory literature from classical decision theory
through to the potentially useful technology of decision analysis and the more recent
contributions of behavioural decision theory. The current debates in the decision
theory literature are then reviewed and this indicates the need for a study that
investigates the relationship between use of decision analysis in investment appraisal
decision2making and organisational performance. In section 9.?, a hypothesis is
advanced for empirical testing.
%(7 ,) )?O+U,IO- OF D)*I0IO- ,)ORC
#onsider first the status of systematic reasoning about human action. !ith stylistic
changes the following, written by -aplace in 5=59, could represent an optimistic view
of decision analysis today &Doward, 56== p@76':
H:y this theory, we learn to appreciate precisely what a sound mind feels
through a kind of intuition often without realising it. The theory leaves
nothing arbitrary in choosing opinions or in making decisions, and we can
always select, with the help of this theory, the most advantageous choice on
our own. It is a refreshing supplement to the ignorance and feebleness of the
human mind.
If we consider the analytic methods brought out by this theory, the truth of its
basic principles, the fine and delicate logic called for in solving problems, the
establishments of public utility that rest on this theory, and its e*tension in the
past and future by its application to the most important problems of natural
philosophy and moral science, and if we observe that even when dealing with
9>
things that cannot be sub8ected to this calculus, the theory gives the surest
insight that can guide us in our 8udgement and teaches us to keep ourselves
from the illusions that often mislead us, we will then realise that there is no
other science that is more worthy of our meditation.K
The possibility of effective, systematic reasoning about human action has been
appreciated for over two hundred years. -aplaceBs predecessor, :ayes, showed in
57@; that probability had epistemological power that transcended its aleatory uses
&Doward, 56=='. In the early 57>>s, :ernoulli captured attitudes towards risk taking
in mathematical form. In his #rs Con$ectandi &575;', ,acob :ernoulli proposed an
alternative to the ob8ectivist view that probability is a physical concept such as a
limiting frequency or a ratio of physically described possibilities. De suggested that
probability is a Hdegree of confidenceK 2 later writers use degree of belief 2 that an
individual attaches to an uncertain event, and that this degree depends on the
individualBs knowledge and can vary from individual to individual. Similarly, -aplace
himself stated in # %hilosophical &ssay of %ro'a'ilities &5=59', that probability is but
the He*pression of manBs ignoranceK and probability calculus is relevant to Hthe most
important questions of lifeK and not 8ust to repetitive games of chance as previously
thought. In addition, Augustus .e +organ in his (ormal Lo!ic &5=<7' argued that:
H:y degree of probability we really mean, or ought to mean, degree of
beliefK &$aiffa, 56@= p97?'
The resurgence of the field in modern times began with statistical decision theory and
a new appreciation of the :ayesian perspective &Doward, 56==' which seeks to
introduce intuitive 8udgements and feelings directly into the formal analysis &$aiffa,
56@='. In his # )reatise on %ro'a'ility &5695' 4eynes took the position that a
probability e*presses the rational degree of belief that should hold logically between a
set of propositions &taken as given hypotheses' and another proposition &taken as the
conclusion' &$aiffa, 56@='. ,effreys &56;6' and ,aynes &56?@', who worked in the
field of physics rather than in mathematics and statistics, provided an all
encompassing view of probability, not as an artefact, but as a basic way of reasoning
about life, 8ust as had -aplace. ,effreys &56;6' and ,aynes &56?@' developed very
clear ways of relating probabilities to what you know about the world around you,
ways that provide dramatic insights when applied to molecular processes that interest
95
many physicists. Dowever, ,aynes &56?@' also showed that these ideas pay off
handsomely when applied to inference problems in our macroscopic world &Doward,
56=='. "rank $amsey was the first to e*press an operational theory of action based on
the dual intertwining notions of 8udgmental probability and utility. In his essay, )ruth
and %ro'a'ility &569@' $amsey adopted what is now termed the sub8ective or decision
theoretic point of view. To $amsey, probability is not the e*pression of a logical,
rational, or necessary degree of belief, the view held by 4eynes and ,effreys, but
rather an e*pression of a sub8ective degree of belief interpreted as operationally
meaningful in terms of willingness to act &$aiffa, 56@='. .e "inetti in his essay,
(oresi!ht: *ts Lo!ical La+s, *ts Su'$ecti,e Sources originally published in 56;7, like
$amsey, assessed a personBs degree of belief by e*amining his overt betting
behaviour. :y insisting that a series of bets be internally consistent or coherent such
that a shrewd operator cannot make a sure profit or HbookK regardless of which
uncertain event occurs, .e "inetti demonstrated that a personBs degrees of belief J his
sub8ective probability assignments J must satisfy the usual laws of probability &$aiffa,
56@='. 0on /eumann and +orgenstern developed the modern probabilistic theory of
utility in their second edition of )heory of -ames and &conomic .eha,iour published
in 56<7. These authors, however, deal e*clusively, with the canonical probabilitiesF
that is, where each outcome is Hequally likelyK. Avidently, they were unaware of the
work of $amsey &$aiffa, 56@= p97@'. Abraham !ald formulated the basic problem of
statistics as a problem of action. !ald &56@<' analysed the general problem in terms
of a normal form analysis &$aiffa, 56@= p977' and the problem he states reduces to
selecting a best strategy for statistical e*perimentation and action when the true state
of the world is unknown. !ald was primarily concerned with characterising those
strategies for e*perimentation and action that are admissible or efficient for wide
classes of prototypical statistical problems. Although !aldBs accomplishments were
truly impressive, statistical practitioners were left in a quandary because !aldBs
decision theory did not single out a best strategy but a family of admissible strategies,
and in many important statistical problems this family is embarrassingly rich in
possibilities. The practitioner wanted to know where to go from where !ald left off.
Dow should he choose a course of action from the set of admissible contendersI The
feeling of !ald and some of his associates was that while this is an important
problem, it is not really a problem for mathematical statisticsF they felt that there 8ust
99
is no scientific way to make this final choice &$aiffa, 56@= p977'. Dowever, they
were in the minority.
In the early 56?>s, there were many proposals suggesting how a decision2maker
should ob8ectively choose a best strategy from the admissible class. /o sooner did
someone suggest a guiding principle of choice, however, than someone else offered a
simple concrete e*ample showing that this principle was counterintuitive in some
circumstances and therefore the proposed principle could not serve as the long sought
key &$aiffa, 56@='. In 56?<, Savage laid the foundations of modern :ayesian
decision theory. In particular he showed that utilities and sub8ective probabilities
could model the preferences and beliefs of an idealised rational decision2maker facing
a choice between uncertain prospects. At least, they should do, if you accept SavageBs
a*iomatic definition of rationality &"rench, 56=<'. :uilding on SavageBs work,
decision analysis was developed in the 56@>s by Doward $aiffa &$aiffa, 56@=F $aiffa
and Schlaifer, 56@5' and $onald Doward &56@=', and represents an evolution of
decision theory from an abstract mathematical discipline to a potentially useful
technology &foreword by (hillips in )oodwin and !right, 5665'.
Simplistically, decision analysis seeks to introduce intuitive 8udgements and feelings
directly into the formal analysis of a decision problem &$aiffa, 56@='. Its purpose is
to help the decision2maker understand where the balance of their beliefs and
preferences lies and so guide them towards a better informed decision &"rench, 56=6
p5='. The decision analysis approach is distinctive because, for each decision, it
requires inputs such as e*ecutive 8udgement, e*perience and attitudes, along with the
Hhard dataK. The decision problem is then decomposed into a set of smaller problems.
After each smaller problem has been dealt with separately, decision analysis provides
a formal mechanism for integrating the results so that a course of action can be
provisionally selected &)oodwin and !right, 5665 p;'. This has been referred to as
the Hdivide and conquerK orientation of decision analysis &$aiffa, 56@='.
.ecompositional approaches to decision2making have been shown to be superior to
holistic methods in most of the available research &for e*ample, 4leinmunt1 et al.,
566@F Dora et al., 566;F +ac)regor and -ichenstein, 5665F +ac)regor et al., 56==F
Armstrong et al., 567?'. "ischer &5677' argues that decompositional approaches
9;
assist in the definition of the decision problem, allow the decision2maker to consider a
larger number of attributes than is possible holistically and encourage the use of
sensitivity analysis. Dolistic evaluations, he believes, are made on a limited number
of attributes, contain considerable random error and, moreover, are e*tremely difficult
when there are fifty or more possible outcomes. 4leinmunt1 &566>' shares this
perspective. De suggests that the consistency of holistic 8udgements will deteriorate
as the number of possible outcomes increases because of the limits on human
information processing capabilities. !hereas he argues, systematic decomposition
rela*es the information processing demands on the decision2maker reducing the
amount of potential error in human 8udgement. "urthermore, since decompositional
methods provide an Haudit trailK it is possible to use them to produce a defensible
rationale for choosing a particular option. #learly this can be important when
decisions have to be 8ustified to senior staff, colleagues, outside agencies, partners, the
general public, or even to oneself &)oodwin and !right, 5665'.
Since its conception the role of decision analysis has changed. /o longer is it seen as
a method for producing optimal solutions to decision problems. As 4eeney &56=9'
points out:
H.ecision analysis will not solve problems, nor is it intended to do so. Its
purpose is to produce insight and promote creativity to help decision2makers
make better decisions.K &)oodwin and !right, 5665 p<'
This changing perception of decision analysis is also emphasised by (hillips &56=6':
Hdecision theory has now evolved from somewhat abstract mathematical
discipline which when applied was used to help individual decision2makers
arrive at optimal decisions, to a framework for thinking that enables different
perspectives on a problem to be brought together with the result that new
intuitions and higher level perspectives are generated.K &)oodwin and !right,
5665 p<'
Dowever, whilst decision analysis does not produce an optimal solution to a problem,
the results from the analysis can be regarded as HconditionallyK prescriptive which
means that the analysis will show the decision2maker what they should do, given the
8udgements that have been elicited from them during the course of the analysis. The
fundamental assumption underlying this approach is that the decision2maker is
9<
rational &)oodwin and !right, 5665'. !hen a decision2maker acts rationally it means
that they calculate deliberately, choose consistently, and ma*imise, for e*ample, their
e*pected preferenceNutility. #onsistent choice rules out vacillating and erratic
behaviour. If it is assumed that managerial decision2makers want to ma*imise, for
e*ample, their personal preferences, and that they perceive that this will happen
through ma*imising the organisationBs ob8ectives, then it may also be assumed that
such managers will pursue the ma*imisation of the organisationBs performance in
meeting its ob8ectives &Darrison, 566? p=5'. +ore simply, if managers are rewarded
based on the organisationBs performance and they behave rationally, they will try to
ma*imise the outcome of their decisions for the organisation, to achieve the highest
amount of personal utility.
"or many years it was believed, implicitly or e*plicitly, that such normative theories
of decision2making not only represent the HoughtK but also the HisK: the normative and
descriptive facets were assumed to be one and the same &4eren, 566@'. The
unprecedented advancements in the physical sciences and information theory and the
realisation of the enormous capabilities inherent in computing machines and
information technology, strengthened and encouraged the belief in rational agents who
were considered to be in full control of their thoughts and actions, and capable of
following the normative desiderata. .ecision failures were e*clusively attributed to
the perceptual2cognitive machine and could, it was assumed, be avoided by increasing
mental effort and by appropriate training &4eren, 566@'. #onsequently, the
presupposition that normative models &with, conceivably, some minor modifications'
can concurrently serve descriptive accounts was introduced with little contention
&4eren, 566@'. "or e*ample, in a frequently quoted article, (eterson and :each &56@7
p96' concluded that:
HIn general, the results indicate that probability theory and statistics can be
used as the basis of psychological models that integrate and account for human
performance in a wide range of inferential tasks.K
There was little attempt to e*plain human behaviour &4eren, 566@'. Aven the most
transparent cases of discrepancy between human behaviour and normative models &for
e*ample, see the often referred to AllaisB parado* outlined in )oodwin and !right,
9?
5665 pp=;2=?' did not change the dominating outlook &4eren, 566@'. In 56?<, !ard
Adwards published his seminal paper HThe Theory of .ecision2+akingK which
marked the birth of behavioural decision theory. Since then, the past forty years have
witnessed a gradual transition in which the descriptive facet has received growing
attention &4eren, 566@'.
:ehavioural decision theory questioned the assumption of normative models that
decisions are, and ought to be, made on solely rational grounds &-ipshit1 and Strauss,
5667'. Such an assumption means that non2cognitive factors such as emotions,
motivations, or moral considerations should have no impact on the decision process
unless they can be 8ustified by rational means. :oth causal observations as well as
growing empirical evidence suggest that this assumption is irreconcilable with any
tenable behavioural descriptive theory &4eren, 566@'. +uch of this research, under
the heading of Hheuristics and biasesK, has portrayed decision2makers as imperfect
information processing systems that are prone to different types of error. The most
pertinent of these studies can be grouped under the headings of probability and
preference assessment and are discussed below.
%ro'a'ility assessments 2 As indicated above, decision analysis, and many other
normative models in decision theory, rely on the use of probability for modelling
the uncertainty surrounding future outcomes. #onsiderable work has been done
on the assessment of sub8ective probabilities, although much of it has focused on
the internal consistency of human assessment &#lemen, 5666'. "or e*ample,
articles in the volume by 4ahneman, Slovic and Tversky &56=9' emphasise how
heuristic 8udgement processes lead to cognitive biases. "or the most part, this
work indicates ways that human 8udgement of sub8ective probability is
inconsistent with probability laws and definitions &#lemen, 5666'. This is a
situation that is e*acerbated in organisational decision2making since many
8udgements are generated by groups of e*perts &#lemen, 5666'. +yers and -amm
&567?' report evidence that face2to2face intervention in groups working on
probability 8udgements may lead to social pressures that are unrelated to group
membersB knowledge and abilities. )ustafson et al. &567;', "ischer &567?',
)ough &567?' and Seaver &567=' all found in their e*periments that interaction of
any kind among e*perts led to increased overconfidence and, hence, worse
9@
calibration of group probability 8udgements. +ore recently, Argote, Seabrigh and
.yer &56=@' found that groups use certain types of heuristics more than
individuals, presumably leading to more biases &#lemen, 5666'.
The situation outlined above is aggravated by the observation that whilst most
people find it easiest to e*press probabilities qualitatively, using words and
phrases such as HcredibleK, HlikelyK or He*tremely improbableK, there is evidence
that different people associate markedly different numerical probabilities with
these phrases &for e*ample, :udescu and !allsten, 566?'. It also appears that, for
each person, the probabilities associated with each word or phrase varies with the
semantic conte*t in which it is used &+organ and Denrion, 566>' and that verbal,
numerical and different numerical e*pressions of identical uncertainties are
processed differently &)igeren1er, 5665F Pimmer, 56=;'. Dence, in most cases
such words and phrases are unreliable as a response mode for probability
assessment &#lemen, 5666'. )iven this, many writers have proposed encoding
techniques. Dowever, the results of the considerable number of empirical
comparisons of various encoding techniques do not show great consistency, and
the articles reviewed provide little consensus about which to recommend &#lemen,
5666'. As +eehl &567= p=;5' succinctly comments:
Hthere are many areas of both practical and theoretical inference in which
nobody knows how to calculate a numerical probability value.K
The most unequivocal result of e*perimental studies of probability encoding has
been that most assessors are poorly calibratedF in most cases they are
overconfident, assigning probabilities that are nearer certainty than is warranted by
their revealed knowledge &+organ and Denrion, 566>'. Such probability
8udgements, -ichenstein, "ischoff and (hillips &56=9' found, are not likely to be
close to the actual long run frequency of outcomes.
Some researchers have investigated whether using specific procedures can improve
probability 8udgements. Stael val Dolstein &5675a and 5675b' and Schafer and
:orcherding &567;' provide evidence that short and simple training procedures can
increase the accuracy &calibration' of assessed probability, although their empirical
results do not indicate an overwhelming improvement in performance. "ischoff
97
&56=9' discusses debiasing techniques intended to improve the quality of sub8ective
performance assessments. )igeren1er and Doffrage &566?' emphasise that framing
8udgements in frequency terms &as opposed to the more traditional sub8ective
Hdegree of beliefK' can reduce assessment bias in a variety of situations. %ther
studies &#lemen, ,ones and !inkler, 566@F Dora, .odd and Dora, 566;' suggest
that embracing the divide and conquer orientation of decision analysis in
probability assessment can improve assessment performance &#lemen, 5666'.
%reference assessment 2 !hile probability assessments can be evaluated readily,
the study of preference and preference assessment techniques, is more problematic
&#lemen, 5666'. The most popular approach to studying preferences has been to
consider the e*tent to which e*pressed preferences are internally consistent, as
e*emplified by the Allais parado* &Allais and Dagen, 5676F Allais, 56?;' or by
Tversky and 4ahnemanBs &56=5' work on framing &#lemen, 5666'. .ecision
analysis prescribes a number of approaches that are formally equivalent for
assessing preference functions &#lemen, 5666'. "arquhar &56=<' surveys many of
the available preference assessment methods. Dershey, 4unreuther and
Schoemaker &56=9' discuss the biases induced by different preference elicitation
approaches in spite of formal equivalence. "ischer &567?' reviews early studies on
the validation of multi2attribute assessment. The typical approach has involved
what is called Hconvergent validityK, which is measured in this case by calculating
the correlation between the intuitive rankings of the sub8ects and the rankings
produced by the preference function &#lemen, 5666'. Although most preference
studies have been aimed at understanding and reducing internal inconsistencies,
4imbrough and !eber &566<' describe an e*periment with a slightly different
orientation. They compared a variety of preference elicitation approaches, each
one implemented via a computer program. Some approaches confronted sub8ects
with their inconsistencies and forced them to make modificationsF these methods
produced recommendations and preference functions that were, by implication,
more acceptable to the users &#lemen, 5666'.
#learly then the research conducted to date in behavioural decision theory has
focussed on the psychology of 8udgement. Since decision analysis is based on a
system of a*ioms, it has been reasonable to study whether people naturally follow the
9=
logic on which decision analysis rests &#lemen, 5666'. Studies have shown that they
do not. "ollowing such observations, there is a tendency in the decision theory
literature for decision analysts and behavioural decision theorists to become
embroiled in a somewhat circular argument over the use and benefits of decision
analysis &for e*ample, see the e*changes between "rench and Tocher summarised in
"rench, 56=6 pp5;625?;'. :ehavioural decision theorists argue that people do not
behave in the manner suggested by decision analysis. .ecision analysts reiterate that
it is not their aim to predict what the decision2maker will do, but rather to suggest to
the decision2maker what they ought to do, if the decision2maker wishes to be
consistent. To behavioural theorists this argument is weak. Tocher &567@ reprinted in
"rench, 56=6 p5;6' writes:
Hany theory which is worth using predicts how people will behave, not how
they should, so we can do our mathematics.K
$ecently researchers such as #lemen and 4wit &9>>>' have attempted to circumvent
this discussion by focussing not on whether people naturally follow the a*ioms of
decision analysis, but on whether learning to do so can lead them to better choices and
consequences.
The relationship between performance and the investment decision2making process
has attracted much theoretical attention &for e*ample, :ailey et al., in pressF Simpson
et al., 9>>>F !ensley, 5666 and 5667F +c#unn, 566=F %tely, 5667F /utt, 5667'. In
5677 Dambrick and Snow advanced a model of interaction between current and past
performance and the investment decision2making process, but concluded that the
effects of the investment decision2making process on performance were not well
articulated and that the available evidence was insufficient to support specific theories
&(apadakis, 566='. Although many other studies &for e*ample, .ean and Sharfman,
566@F Dart, 5669F Guinn, 56=>' have described and e*plained the investment
decision2making process, little consensus has emerged as to the e*pected relationship
between organisational performance and investment decision2making processes &for
e*ample, (riem et al., 566?F $a8agopalan et al., 566;'. Specifically, whilst it is well
established that management science and operations research add value to
organisations when used well &#lemen and 4wit, 9>>>', the value of decision analysis
96
remains less well documented. Although many successful applications have been
performed and published &for e*ample, %tis and Schneiderman, 5667F /angea and
Dunt, 5667', the evidence remains largely anecdotal and unsystematic &#lemen and
4wit, 9>>>'. .espite over four decades of research developing decision analysis
techniques, gaining an understanding of the behavioural and psychological aspects of
decision2making, and the application of decision analysis to real organisational
decisions, no research has been able to show conclusively what works and what does
not &#lemen, 5666'. It is highly likely that being unable to document the value of a
decision analysis approach to investment appraisal decision2making has hampered
some proponents as they have tried to gain acceptance for decision analysis within
their organisations &see Section @.; of #hapter @ and #lemen, 5666'. This could be
seen as contributing directly to the gap between current practice and current capability
in investment appraisal. If decision analysis could be shown to be definitively of
value, and that this value easily overwhelms the typical costs of compiling the
modelling and analysis, decision analysis would become much more attractive to
organisations &Section @.; of #hapter @F #lemen, 5666'. #onsequently, in time, the
current gulf between theory and practice would narrow. "urthermore, such research
would contribute to the theoretical debate between decision analysts and behavioural
decision theorists &#lemen, 5666'. If, as many decision theorists believe &for
e*ample, "rench, 56=6', companies that use decision analysis outperform those that
do not, such research would contribute to the theoretical debate between the decision
analysts and behaviouralists. The behavioural decision theorists would no longer be
able to claim that there is no value in a theory that does not aim to predict what
decision2makers will do. The third research question that this thesis aims to e*plore
then, is the question of whether success in decision2making depends on the decision2
making process managers use &Ditt and Tyler, 5665' and, specifically, whether
adopting decision analysis techniques in investment appraisal decision2making has a
positive effect on organisational performance.
The literature reviewed in this section has indicated that there is a need for a study to
investigate the e*istence of a relationship between the use of decision analysis
techniques and concepts in investment appraisal decision2making and organisational
performance. This is the third research question that this thesis aims to answer.
Dowever, before such a link can be proved to e*ist, two assumptions must hold. The
;>
ne*t section begins by stating these assumptions and proving their validity. It
continues to review previous studies that have been undertaken investigating the
relationship between business performance and various aspects of the organisational
investment decision2making process. Specifically, the section focuses on those
studies that have concentrated on the effects of rationality, formality and consensus in
the decision2making process since these are all features inherent in using decision
analysis techniques and concepts. The section concludes by advancing a hypothesis
for empirical testing.
%(8 D)*I0IO- A-A+C0I0 A-D ORGA-I0A,IO-A+ ')RFORMA-*)
As .ean and Sharfman &566@' observe, the following two assumptions must hold to
prove a link between investment decision process and decision effectiveness. "irstly,
it must be assumed that investment decision processes are related to choicesF or, more
specifically, that the investment decision process followed influences the choices
made. Although this assumption appears intuitively obvious, many academics have
argued that the operating environment shapes organisational and individual choices
&for e*ample, Aldrich, 5676F (feffer and Salancik, 567='. %thers, however, claim that
despite the e*istence of these e*ternal factors, managers retain a substantial degree of
control over choices &for e*ample, +iles, 56=9F #hild, 5679'. %ne argument made in
favour of this position by .ean and Sharfman &566@' is that some managers make
very poor choices with devastating consequences for their firms, while others in very
similar circumstances make much better choices &for e*ample, :ourgeois, 56=<'.
Such variation, the authors assert, could not e*ist if constraints alone were driving
decisions. Dence, .ean and Sharfman &566@' conclude that it appears likely that
viable outcomes are a product of the decision process used. -eading on from this, the
second assumption is that choices relate to outcomes, and that all outcomes are not
equally good. %nce again there can be very little doubt that e*ternal forces also
influence decision effectiveness &Ditt and Tyler, 5665F (feffer and Salancik, 567='.
#hanges in competitor strategies or customer tastes can turn strategic coups into
disasters or vice versa. Dowever, .ean and Sharfman &566@' note that it is unlikely
that the influence of such forces eliminates the impact of choice on decision
effectiveness as it is hard to imagine a decision in which all potential choices will be
equally successful or unsuccessful.
;5
The two assumptions then appear plausible &.ean and Sharfman, 566@' which
suggests that it is reasonable to e*pect the investment appraisal decision2making
process to influence decision effectiveness. Dowever, as Aldrich rightly observed
&5676', the importance of managerial decisions in determining organisational
outcomes is ultimately an empirical question &.ean and Sharfman, 566@'. +any
empirical studies have investigated the e*istence of a relationship between the
investment decision2making process and effectiveness. /one have concentrated on
the use of decision analysis in the investment decision2making processes of
organisations. Dowever, several have e*plored the effects of comprehensiveness,
rationality, formality and consensus in the decision2making process on organisational
performance. In much of the decision theory literature, it is argued that decision
analysis provides:
Hconvincing rationale for choice, improves communication and permits
direct and separate comparisons of different peopleBs conceptions of the
structure of the problem, and of the assessment of decomposed elements
within their structures, thereby raising consciousness about the root of any
conflict.K &Dumphreys, 56=> in )oodwin and !right, 5665 p577'
)oodwin and !right &5665' also argue that adopting a decision analysis approach
implies comprehensivenessNrationality and formalisation of the decision2making
process, improved communication amongst the stakeholders and provides the
organisation with access to a common language for discussing the elements of a
decision problem. This, they argue, helps to build consensus in the company, which
in turn e*pedites implementation of the decision. 4eeney and $aiffa &5679 pp5>255'
say of decision analysis:
HAs a process, it is intended to force hard thinking about the problem area:
generation of alternatives, anticipation of future contingencies, e*amination of
dynamic secondary effects, and so forth. "urthermore, a good analysis should
illuminate controversy J to find out where basic differences e*ist, in values
and uncertainties, to facilitate compromise, to increase the level of debate and
undercut rhetoric J in short, Hto promote good decision2makingK.K
Since adopting decision analysis clearly involves comprehensiveness, rationality,
increased formality and high levels of organisational consensus, it suffices to e*amine
that empirical literature that has e*amined the relationship between these aspects of
;9
the investment decision2making process and decision effectiveness. These studies are
now e*amined. Attention is first focussed on the effect of comprehensiveness and
rationality in the decision2making process.
Smith et al. &56==' provided some empirical support for a positive relationship
between performance and comprehensivenessNrationality in the decision2making
process. They found that, for both small and larger firms, comprehensive decision2
making processes out2performed less comprehensive. Similarly, ,ones et al. &5669'
reported consistently positive relationships between organisational effectiveness and
comprehensiveness in decision2making. In addition, a series of publications on
hospital integration strategies &for e*ample, :lair et al., 566>', researchers found that
successful ventures were associated with comprehensive strategy formulation
processes &(apadakis, 566='. ,anisB &56=6' case studies suggested that public policy
decisions that used rational methods were more successful than those that did not.
(apadakisB &566=' study also provided evidence that the companies that e*hibit the
strongest organisational performance tend to be those with rational decision2making
processes, a participative approach and e*tensive financial reporting. "urthermore,
studies by #apon et al. &566<' and (earce et al. &56=7' suggest that formalisation in
strategic planning is positively related to organisational performance. Such results led
(apadakis &566=' to hypothesise that performance is positively related to
comprehensivenessNrationality and formalisation in the investment decision2making
process.
#onversely, "redrickson and his colleagues &"redrickson and Iaquinto, 56=6F
"redrickson, 56=?F "redrickson, 56=<F "redrickson and +itchell, 56=<' looked at
prototypical &assessed by response to a scenario' rather than actual investment
decision2making processes and related them to firm performance rather than to
specific decision outcomes and concluded that:
H"irms usually do not use slack generated by e*cellent performance to pay the
costs of seeking optimal solutionsF instead resources are absorbed as sub2
optimal decisions are made. This phenomenon may help e*plain why
managers in historically successful firms sometimes make a series of what
appear to be inadequately considered, intuitive decisions that in combination
have significant negative consequences.K &"redrickson, 56=? p=9<'.
;;
Similarly, #yert and +arch &56@;' argued that superior performance lowered the
intensity with which organisations HsearchedK for and analysed information. +ore
specifically, :ourgeois &56=5' and +arch and Simon &56?=' proposed that slack
resources permit organisations the Hlu*uryK of HsatisficingK and sub2optimal decision2
making. !hereas in poorly performing organisations the lack of basic funds e*erts
pressure on management during the making of crucial decisions, as a wrong decision
may drive the firm out of business. #onsequently, since management has less scope
for error, they may have strong incentives to follow rationalNcomprehensive processes
&:ourgeois and Aisenhardt, 56==F #yert and +arch, 56@;'. This suggests that
managers of poorly performing firms may hire consultants, seek advice from various
sources and conduct e*tensive financial analyses &(apadakis, 566='. Such
observations led "redrickson &56=?' to conclude that the investment decision2making
process of poor performers is more comprehensive than that of e*cellent performers.
The above arguments, if correct, would indicate that good organisational performance
is negatively related to comprehensivenessNrationality in the investment decision2
making process &(apadakis, 566='.
#learly, then, much of the research to date appears to have produced contradictory
results and no consensus seems to have yet emerged. #ontrary to the arguments of
"redrickson &56=?' and others, it can be argued that good performance enables
companies to rationaliseNmodernise their internal structure and systems and thus be in
a position to apply more rationalNcomprehensive and formalised investment decision2
making processes for two reasons. "irstly, as .ean and Sharfman &566@' have
previously argued, effective decisions must be based on organisational goal. $ational
decisions usually require e*tensive data collection and analysis efforts and it is
difficult to do this unless the decision is closely aligned to the organisationsB
ob8ectives &-angley, 56=6'. Ditt and Tyler &5665, p;96' described rational, formalised
decision2making as a series of analytical processes in which a set of ob8ective criteria
is used to evaluate strategic alternatives. This orientation toward organisational goals
makes it more likely that procedurally rational decisions will be effective &.ean and
Sharfman, 566@'. Secondly, formalised, rational decisions are also likely to involve
relatively complete information and knowledge of constraints. A*ecutives who
collect e*tensive information before making decisions will have more accurate
;<
perceptions of environmental conditions, which has been shown to relate positively to
firm performance &:ourgeois, 56=?'.
A second stream of research deals with the impact of consensus on performance.
.espite the profound importance given to the performance2consensus relationship in
the normative literature &(apadakis, 566=', there is still much disagreement in the
empirical literature which indicates that more testing is required &Akenberg, 9>>>F
(riem, 566>F .ess, 56=7F .ess and %riger, 56=7'. #onsensus is the agreement of all
parties to a group decision &(apadakis, 566='. #urrent thinking attributes tremendous
significance to the homogenisation of perceptions and to goal consensus, which is
assumed to be fundamental to good economic performance &(apadakis, 566=F
:ourgeois, 56=?'. #hild &567<' was among the first to propose that homogeneity
among the members of the top management team as to the ob8ectives contributes to
higher performance &(apadakis, 566='. Similarly, :ourgeois &56=5' argued that the
organisational slack generated by business success functions as a source of conflict
resolution. Since when a company is on a Hwinning trackK &(apadakis, 566='
everyone prefers to be associated with the winner and there is less place for political
activities and long debates over goals and priorities &.ess, 56=7'. A number of
empirical studies have confirmed the e*istence of a positive relationship between
organisational performance and consensus. "or e*ample, Aisenhardt and :ourgeois
conducted several studies on decision2making in dynamic environments. The results
from one of their studies suggested that political behaviour within top management
teams leads to poor organisational performance &Aisenhardt and :ourgeois, 56=='.
Dowever, the ma8ority of studies in this area have been conducted in the laboratory,
where environmental forces are not an issue, and the few field studies that have been
carried out have not attempted to assess actual decision outcomes &Schweiger et al.,
56=@'. The e*ception is .ean and SharfmanBs 566@ study of twenty2four companies
in si*teen industries, which provided an indication that the decision process that was
followed influenced the decision2making effectiveness. 3nlike earlier studies, the
researchers included environmental factors and the quality of implementation of the
decision in their model. %ne of their main findings was that managers who engaged
in the use of power or pushed hidden agendas were less effective than were those who
did not. %ther studies by ,anis &56=6', "ord &56=6' and /utt &566;' have all indicated
a link between politics and unsuccessful decisions.
;?
Dowever, conversely, some researchers have provided evidence that too much internal
consensus may be dysfunctional. "or e*ample, !hitney and Smith &56=;' argued that
an emphasis on organisational or management consensus could reduce individualsB
receptivity to information that contradicts the views of the dominant coalition despite
the fact that such information may be vital for the quality of the final decision. Thus,
the pressure for consensus postulated by normative methods to decision2making may
produce negative results &(apadakis, 566='. Investigating the performance2consensus
relationship, )rinyer and /orburn &567727=' found that the highest performing firms
e*perienced a negative correlation between performance and consensus. Thus they
hypothesised that high levels of cohesiveness may be dysfunctional, and that some
disagreement among members of the top management team may be an internal
strength related to superior performance &(apadakis, 566='. -angley &566?' also
warned that when everyone in power instinctively shares the same opinion on an
issue, the wise manager should be wary. 3nanimity, she writes, is unlikely to lead to
an ob8ective evaluation of options, and normal checks and balances may be short2
circuited. -angley argues that unanimity may mean that a proposal has strong value,
but it may also be symptomatic of a disturbing trend, that is, a uniformity in which
members share values and beliefs and that e*cludes deviation from the decision2
making process. She concludes that whilst obviously a strong culture has many
advantages, when the organisation is faced with discontinuities this same culture
becomes a liability as common beliefs become invalid.
"inally, contrary to both the above streams of results, !ooldridge and "loyd &566>'
found no statistically significant relationship between consensus and organisational
performance.
Avidently then, the performance2consensus research has produced some conflicting
results. This may be attributed to differences in units of analyses, in methodologies
and research questions &.ess and %riger, 56=7', or, perhaps, even to the nature and
stage of the strategic process under investigation which may impact upon the scope,
content and degree of consensus &!ooldridge and "loyd, 566>F (apadakis, 566='.
+ore interestingly, (apadakis &566=' postulates that a lack of any significant
relationship suggests the co2e*istence of two opposite effects that Hcancel each other
;@
outK in practice. .ean and Sharfman &566@' have argued that effective decisions must
be based on organisational goals. (olitical decision processes are, by their very
nature, organised around the self2interests of individuals or groups &(feffer, 56=5F
(ettigrew, 567;', which are often in conflict with those of the organisation.
Therefore, it can be argued that good performers are less likely to e*hibit less politics
and less problem2solving disagreement in their decision2making process.
This section has 8ustified the assumptions that must hold in order to prove a link
between investment decision process and effectiveness. It has reviewed those
empirical studies that have focussed on the effects of comprehensiveness, rationality,
formality and consensus in the decision2making process on organisational
performance. It has provided evidence that using decision analysis means rationality,
comprehensiveness, formality and increased consensus in investment decision2
making. It therefore suffices to advance only one hypothesis for empirical testing in
this thesis: organisational performance is positively related to use of decision analysis
in investment appraisal decision2making. In answering the third research question,
the researcher aims to investigate this proposition.
The current study will use the indication of current capability and current practice
gained from answering the first and second research questions to rank the companies
according to the number of techniques used in their investment appraisal process. The
research will then assume that any value added to the company from using a decision
analysis approach, including any HsoftK benefits, ultimately affects the bottom2line.
This assumption will be 8ustified in #hapter 7. It means that it is therefore
permissible to use publicly available financial data to indicate business success. The
e*istence of a relationship between organisational performance and use of decision
analysis in investment appraisal will then be analysed statistically.
%(: *O-*+U0IO-
In seeking to e*plore the investment decision2making processes of companies, the
literature review for the current study has e*amined the academic literature on
investment decision2making. The source of each of the three research questions
proposed in #hapter 5 was e*plored and a hypothesis advanced for empirical testing.
;7
The ne*t chapter e*amines the conte*t for the current study. It will show how the oil
and gas industry is such an e*treme e*ample of investment appraisal decision2making
under conditions of risk and uncertainty that it provides a useful environment in which
to study investment decision2making.
;=
*hapter ,hree
,he Oil Industry in the U(.(
;6
5(3 Introduction
This chapter draws on the oil industry literature to present a brief description of the
industry that highlights the main challenges facing it in the 95
st
century. Since the
current study focuses on oil and gas companies that operate in 3.4., the effects of
these global changes on the 3.4. industry are e*amined. This indicates the growing
comple*ity of the business environment of those companies operating in the upstream
oil and gas sector and highlights why decision analysis is beginning to receive
increasing attention in the industry and, consequently, why it provides such a useful
conte*t in which to study investment decision2making.
5(% *urrent challenges in the glo#al oil industry
(or o,er a century and a half, oil has 'rou!ht out 'oth the 'est and
+orst of our ci,ilisation. *t has 'een 'oth 'oon and 'urden. &ner!y is
the 'asis of our industrial society. #nd of all ener!y sources / oil has
loomed the lar!est and the most pro'lematic 'ecause of its central role,
its strate!ic character, its !eo!raphic distri'ution, the recurrent pattern
of crisis in its supply / and the ine,ita'le and irresisti'le temptation to
!rasp for its re+ards. *ts history has 'een a panorama of triumphs and
a litany of tra!ic and costly mistakes. *t has 'een a theatre for the no'le
and the 'ase in the human character. Creati,ity, dedication,
entrepreneurship, in!enuity, and technical inno,ation ha,e coe0isted
+ith a,arice, corruption, 'lind political am'ition, and 'rute force. 1il
has helped to make possi'le mastery o,er the physical +orld. *t has
!i,en us our daily life and, literally, throu!h a!ricultural chemicals and
transportation, primacy. *t has also fuelled the !lo'al stru!!les for
political and economic primacy. 2uch 'lood has 'een spilled in its
name. )he fierce and sometimes ,iolent "uest for oil / and for the
riches and po+er it con,eys / +ill surely continue so lon! as oil holds a
central place since e,ery facet of our ci,ilisation has 'een transformed
'y the modern and mesmerisin! alchemy of petroleum.
<>
The above paragraph has been adapted from the closing remarks made by .aniel
Cergin in his book, )he %rize &5665', which chronicles the development of the worldBs
oil industry. Three themes are used to structure the book and these clearly illustrate
the global impact of the oil and gas industry. The first of these is that oil is a
commodity intimately intertwined with national strategies, global politics and power
as evidenced by its crucial role in every ma8or war in the last century. The second is
the rise and development of capitalism and modern business. According to Cergin
&5665 p5;':
H%il is the worldBs biggest and most pervasive business, the greatest of the
great industries that arose in the last decades of the nineteenth century.K
A third theme in the history of oil illuminates how ours has become a Hhydrocarbon
societyK &Cergin, 5665 p5<'. %il has become the basis of the great post2war sub2
urbanisation movement that transformed both the contemporary landscape and our
modern way of life. Today, it is oil that makes possible, for e*ample, where we live,
how we live, how we commute to work and how we travel as well as being an
essential component in the fertiliser on which world agriculture depends, and key
material in the production of pharmaceuticals.
)lobally, the industry has evolved from primitive origins through two world wars, the
Sue1 #anal crisis, the )ulf !ar and significant fluctuations in supply and demand, all
with their subsequent impact on the oil price, to become a multi2billion pound
business comprised of some of the worldBs biggest and most powerful companies. It
is now recognised as an essential national power, a ma8or factor in world economies, a
critical focus for war and conflict, and a decisive force in international affairs &Cergin,
5665 p776'. Dowever, the global industry is changing. "our factors in particular are
contributing to the uncertainty surrounding the industryBs future. These are reviewed
in this section. The following section analyses the effect of these challenges on the
3.4. oil and gas industry. The impact of these recent changes on investment decision2
making in the industry will then be discussed.
<5
"ield si1e
)lobally many of the oil ma8ors still generate much of their output J and profits J
from giant fields discovered decades ago. "or e*ample, in 566@ it was estimated that
=>E of :(Bs &:ritish (etroleum' oil and gas production was from /orth America and
:ritain, mainly from a handful of large fields in Alaska and the /orth Sea &The
Aconomist, 566@'. (roduction from nearly all these giant fields is either near its peak
or is already declining. /ew fields are rarely as large or as profitable as these earlier
large reservoirs. !orldwide since the mid256=>s, few giant oilfields have been
discovered &figure ;.5' and, although, many smaller fields have been found, they have
not delivered the same economies of scale &The Aconomist, 566@'.
"igure ;.5: 3orld+ide !iant fields (initial reser,es 'y disco,ery year (source: Camp'ell, 1997 p45
&
8
3&
38
%&
%8
5&
58
7&
78
367& 3678 368& 3688 36:& 36:8 369& 3698 36;& 36;8 366& 3668
1illion 1arrels
!8yr moving
average$
<9
"inite resource
!hilst virtually everyone is agreed that oil is a finite resource there is much
disagreement in the industry about e*actly when demand will irreversibly e*ceed
supply. Some analysts, such as #ampbell &5667' argue that production of
conventional oil, which he defines to be that oil with a depletion pattern which starts
at 1ero, rises rapidly to a peak, and then declines rapidly, will peak in 9>5> &figure
;.9'. %thers believe it will last much longer:
Hthe world is running into oil not out of it ... The issue Lof limited oil
resources will beM unimportant to the oil market for ?> yearsK &%dell, 566?'
(i!ure 6.5: Camp'ell7s prediction of +orld oil production after 1998 (actual production to 1998 and
then predicted thereafter (source: Camp'ell, 1997 p199
Cet, much of this is con8ecture. !hat is known is that worldwide proven reserves
have increased by appro*imately two thirds since 567> but the countries that contain
ample quantities of low cost oil, and which account for most of that increase, are
currently inaccessible to western firms &figure ;.;'. +iddle Aastern countries that are
members of the %rganisation of (etroleum A*porting #ountries &%(A#', for e*ample,
account for almost si*ty percent of the worldsB proven reserves &figure ;.<'.
World Oil Production (after Campbell 1997)
&
8
3&
38
%&
%8
5&
367& 36:& 36;& %&&& %&%& %&7& %&:&
1illions o"
#arrelsDyear
<;
"igure ;.;: :istri'ution of remainin! (;et-to-%roduce oil (in .illions of .'ls 'y country (calculated
'y su'tractin! total production of con,entional oil to date from Camp'ell7s estimate of cumulati,e
production of con,entional oil and di,idin! 'y country (source: Camp'ell, 1997 p 94
"igure ;.<: :istri'ution of remainin! (;et-to-%roduce 1il (in .illions of .'ls 'y re!ion (calculated
'y su'tractin! total production of con,entional oil to date from Camp'ell7s estimate of cumulati,e
production of con,entional oil and di,idin! 'y re!ion (source: Camp'ell, 1997 p94
Dowever, nationalism runs high in Saudi Arabia and 4uwait and relationships within
Iraq are tenuous J a situation which is unlikely to change in the near future &The
Aconomist, 566@'. +oreover, whilst the statistics might indicate that technically the
oil firms are reporting increased reserves in reality this conceals two trends. "irstly,
&
3&&
%&&
5&&
7&&
8&&
:&&
M
i
d
d
l
e

)
a
s
t
)
u
r
a
s
i
a
+
a
t
i
n
A
m
e
r
i
c
a
A
"
r
i
c
a
-
o
r
t
h
A
m
e
r
i
c
a
/
e
s
t

)
u
r
o
p
e
)
a
s
t
M
i
d
d
l
e

)
a
s
t
!
o
t
h
e
r
$
!
U
n
"
o
r
s
e
e
n
$
Remaining yet-
to-produce
oil in 1illions
o" 1#ls

&
8&
3&&
38&
%&&
%8&
0
a
u
d
i
a

A
r
a
#
i
a
F
o
r
m
e
r

0
o
v
i
e
t

U
n
i
o
n
I
r
a
@
I
r
a
n
A
#
u

D
h
a
#
i
.
u
2
a
i
t
U
0
A
?
e
n
e
E
u
e
l
a
*
h
i
n
a
M
e
=
i
c
o
+
i
#
y
a
-
i
g
e
r
i
a
-
o
r
2
a
y
U
.
A
l
g
e
r
i
a
Remaining
yet-to-produce
oil in 1illions
o" 1#ls
<<
by using new technology either to e*tend field life or to e*ploit fields that were
previously inaccessible, oil companies have been able to increase their reported
reserves. Secondly, petroleum companies are becoming increasingly reliant on gas
which is harder to transport and less profitable to produce &The Aconomist, 566@'.
Some, such as -aherrQre &5666', are more cynical and believe that the bulk of the
recent Hreserves growthK can be attributed to faulty reporting practices.
.emand
!orld demand for oil, gas and coal in the 95
st
century will depend on two contrary
forces. "irstly, there is the possible reduction in demand by the countries in the
%rganisation for Aconomic #o2operation and .evelopment &%A#.' caused by
structural changes, saturated markets, ageing populations and increasing efficiency.
Such efficiency gains are driven by competition, concerns for energy security and
environmental measures. Action to meet 4yoto targets, set in a summit on global
warming in 4yoto, ,apan in .ecember 5667, will put a cost on carbon emissions J
either by ta*ation or by trading. #oal and oil will face fierce competition in power
generation. As indicated above, oil ma8ors are relying increasingly on gas &The
Aconomist, 566@'. Skeikh Ahmed Paki Camani believes that new hybrid engines
could cut petrol consumption by almost ;>E, while fuel2cell cars, which he predicts
will be widely used by 9>5>, will cut demand for petrol by 5>>E. In a recent article
in &ner!y :ay he said:
HThirty years from now there will be a huge amount of oil J and no buyers.
%il will be left in the ground. The Stone Age came to an end not because we
had a lack of stones and the oil age will come to an end not because we have a
lack of oil.K &Anergy .ay, ;
rd
,uly 9>>> p7'
Dis claims are substantiated by a study from 3.S. based #llied .usiness *ntelli!ence
&A:I', which forecasts millions of fuel2cell vehicles by 9>5>. A:I business analyst
Atakan %1bek is also quoted in the same &ner!y :ay article:
H:y the second decade of this century mass production of automotive fuel
cells will result in first a glut in the world oil supply and then in a total
reduction of oil as a vehicle fuel.K &Anergy .ay, ;
rd
,uly 9>>> p7'
<?
Secondly, there is the potential demand in developing countries. Dow it is fulfilled
depends on future economic growth. The oil companies, however, are optimistic with
Shell suggesting that energy consumption will be between si*ty and eighty percent
higher by 9>9>, with developing countries consuming over half of the available
energy &+oody2Stuart, 5666'.
$estructuring
International oil prices are notoriously volatile &figure ;.?'. Dowever, when, in the
winter of 566=25666, oil prices dropped to their lowest levels in real terms for twenty2
five years, the profit margins of even the largest companies were squee1ed and all
companies were forced to reduce costs. This proved difficult and with the need to
improve their return on capital employed, which has historically been lower than the
cost of that capital, the boards of some of the largest companies perceived the only
way to make further savings was through big mergers, followed by ruthless
restructuring &The Aconomist, 566='.
"igure ;.? : #ctual spot .rent oil price o,er time (source: .% Statistical <e,ie+ of 3orld &ner!y, 5999
In 566= :( agreed to buy Amoco for R<= billion, A**on and +obil, AmericaBs biggest
oil firms, announced a R77 billion merger that has made A**on +obil the worldBs
biggest oil firm J and, on some measures, the largest firm in the world. The merger is
&
8
3&
38
%&
%8
5&
58
36;5 36;8 36;9 36;6 3663 3665 3668 3669 3666 %&&&

FD1arrel
<@
already starting to transform the worldBs oil industry. "irms that were once considered
big, such as #hevron and Te*aco, are rushing to find partners. This is true even in
Aurope, where national champions have traditionally resisted pressures to merge.
"ranceBs Total announced in 566= that it was buying :elgiumBs (etrofina for some
R5; billion &The Aconomist, 566=' and, more recently, Total "ina have also bought
"ranceBs Alf.
!hilst some argue that this is 8ust typical oil industry over2reaction to the bottom of
the price cycle &for e*ample, Auan :aird of Schlumberger in The Aconomist, 566=',
others believe that the structure of the oil industry has altered irreversibly:
Hthe changes unleashed by the mergers look unstoppableK &The Aconomist,
566= p7<'
Indeed, whilst there may well always be a role for the Hscrappy entrepreneurK &The
Aconomist, 566=', si1e is becoming increasingly important in the oil industry. It takes
a great deal of capital and a Hmatching appetite for riskK &The Aconomist, 566=', to
succeed in the #aspian or !est Africa. Tackling a R@ billion pro8ect in #hina will be a
huge effort for Te*aco, with its revenues of some R?> billion. "or A**on +obil
though, which is four times that si1e, such pro8ects will be, according to The
Aconomist &566=', Hsmall potatoesK.
This section has highlighted the current global challenges facing the oil industry.
Since the current study will focus on those petroleum companies operating in the
3.4., the ne*t section e*amines the effect of the worldwide challenges on the 3.4.
industry. The impact on investment decision2making will then be investigated.
5(5 ,) OI+ I-DU0,RC I- ,) U(.(
In the 3.4. there are appro*imately 9?7 offshore fields currently in production on the
3nited 4ingdom #ontinental Shelf &34#S' and 59 under development. In 5666 in
the 3.4. /orth Sea, daily oil output averaged 9.@6 million barrels per day including a
contribution of some =6,>>> barrels per day from onshore fields. In 9>>>, !ood
+acken1ie predicts that oil production will remain at this level. In total, /orth Sea
<7
production &including /orway' averaged some @.5? million barrels per day in 5666.
This is forecast to increase to an average of some @.<@ million barrels per day in 9>>>
&!ood +acken1ie /ewsletter, "ebruary 9>>>'. Since 56@<, the industry has
contributed significantly to the 3.4. economy. It has provided, via ta*es, S=6 billion
to the e*chequerF significant employment, with currently ;>,>>> 8obs offshore and
over ;>>,>>> direct and indirect 8obs onshore &"oreword of The %il and )as Industry
Task "orce $eport published by the .epartment of Trade and Industry, 5666'F and in
5666 it was responsible for ;@E of the 3.4.Bs industrial investment &3.4. Anergy in
:rief published by the .epartment of Trade and Industry, 9>>>'.
Dowever, in the early 567>s the average si1e of a 34#S discovery was about one
billion barrels of oil &:rown, 5669'. Today, nearly half of all developed fields in the
34#S contain less than fifty million barrels of oil &Shell, 566='. This decline is
shown in figures ;.@ and ;.7.
"igure ;.@: )he a,era!e size of =.>. fields 'y disco,ery year (source: =nited >in!dom 1ffshore
1perators #ssociation, 5999a, http:??+++.ukooa.co.uk
&
%&&
7&&
:&&
;&&
3&&&
3%&&
369& 369% 3697 369: 369; 36;& 36;% 36;7 36;: 36;; 366& 366% 3667
Million #arrels
o" oil e@uivalent
<=
"igure ;.7: :isco,eries 'y field-size class in the @orth Sea (source: Camp'ell, 1997 pAB
/early all the significant discoveries made in the first ten years of /orth Sea
e*ploration have been developed or are approved for development. Typically,
operators in the /orth Sea are focusing on cost cutting strategies as production from
their e*isting fields declines. .ue to high maintenance costs and low margins, large
operators are pursuing retrenchment strategies by selling off their mature oil fields to
smaller lower cost operators in order to concentrate their resources &human and
financial' on much larger and rewarding fields in other countries. $eflecting this
trend, over the past two years e*ploration activity on the 34#S has declined
substantially &The %il and )as Industry Task "orce $eport published by the
.epartment of Trade and Industry, 5666'.
In the 34#S, e*perts estimate that there is a further si*teen to twenty billion barrels
of oil equivalent remaining in fields that are either already in production or under
development and between five and thirty billion barrels of oil equivalent yet to be
found &3nited 4ingdom %ffshore %perators Association, 9>>>b'. Dowever, much of
this is likely to be located in smaller, increasingly comple* geological structures
requiring innovative techniques to develop them safely and viably. The costs of
developing fields and producing oil from the 34#S remain higher than other oil
basins with similar characteristics elsewhere in the world &for e*ample, see figure
;.='. .ownward pressure on oil prices has coincided with the rising costs of
operations for new field developments. Increasing competition for new investment is
coming from other prospective areas of the world, where countries are now offering
competitive fiscal and regulatory terms and conditions. In 5666, a task force team of
&
%
7
:
;
3&
3%
37
3:
36:8 369& 3698 36;& 36;8 366& 3668
*umulative
discoveries
G31#l
666-8&&M#l
766-3&&M#l
H3&&M#l
<6
e*perts from the 3.4. )overnment and oil industry considered the current position
and future scenarios for the industry in the 3.4.. Their conclusions were that total
34#S production is likely to peak at over si* million barrels of oil equivalent per day
in the near future. .ue to the maturity of the /orth Sea as an oil province, production
will then start to fall, even allowing for the potential reserves in the Atlantic +argin
&The %il and )as Industry Task "orce $eport published by the .epartment of Trade
and Industry, 5666'. Dowever, they perceived the speed of this decline to be
dependent on a number of factors, including: technology, the cost base of activity, the
level of continued investment 2 particularly in e*ploration, the maintenance of
e*isting infrastructure, the fiscal regime and the level of world oil prices.
"igure ;.=: 3orld+ide operatin! costs (source: =nited >in!dom 1ffshore 1perators #ssociation,
5999c, http:??+++.ukooa.co.uk
In the 3.4., these pressures have led the 3.4. )overnment and industry to seek ways
to improve the attractiveness of the 3.4. in comparison to international petroleum
provinces. The 3.4. )overnment introduced the first out2of2round licensing awards
in 5669 to facilitate the early development of fields &.epartment of Trade and
Industry, 566='. In the same year, the #$I/A &#ost $eduction In the /ew Ara'
/etwork was launched. This is an initiative supported by the 3.4. oil and gas
e*ploration and production industry that aims to increase the global competitiveness
of the 3.4. by focussing on reducing the costs of its participants 2 the domestic and
,
h
a
i
l
a
n
d
M
a
l
a
y
s
i
a
I
n
d
o
n
e
s
i
a
A
n
g
o
l
a
A
u
s
t
r
a
l
i
a
*
o
l
u
m
#
i
a
*
o
n
g
o
.
a
E
a
k
h
s
t
a
n
U
.
-
o
r
2
a
y
DollarD1arrel
oil e@uivalent
?>
international operators that operate on the 34#S and the contracting and supply
companies in the 3.4.. The resulting cost reduction initiative has led to savings of
;>E in capital and operating costs &.epartment of Trade and Industry, 566= p<'. In
566@, #$I/A turned its attention to competitiveness in a wider sense than cost
reduction. The ob8ective became to find ways of enhancing the value of the services
and equipment provided by contractors and suppliers to operators not 8ust in field
developments but also in field operations. The aim is to e*tend the commercial life of
the 34#S and through improving the global competitiveness of the supply industry,
to increase e*port market share and to secure employment in this sector well beyond
the time when 3.4. becomes a net importer of oil again. In response, in 5667 the
.epartment of Trade and IndustryBs Infrastructure and Anergy (ro8ects .irectorate
commissioned a study on the impact of changing supply chain relations on the
upstream oil and gas supplies industry particularly on small, medium2si1ed
enterprises. At the time of the study, supply chain management &S#+' was seen as a
set of management processes that could usefully help the industry cope with some of
the challenges it then faced. !hen the oil price dropped in 566=, what was simply
desirable became imperative. S#+ is now seen as a technique that can make a
considerable contribution to the economics of upstream development. The #$I/A
network S#+ initiative has identified cost savings and substantial improvements in
value through the implementation of a range of S#+ and collaborative initiatives. In
parallel with the #$I/A network S#+ initiative, the %il and )as Industry Task "orce
&%)IT"' has been established. The %)IT" was launched in /ovember 566= to
address the impact on the 3.4.Bs oil and gas industry of the low world oil price. The
resulting initiatives were announced in September 5666 and include the creation of a
new industry body -%)I# &-eading %il and )as Industry #ompetitiveness' which
will seek to change fundamentally the way in which the oil and gas industry does
business by driving improved S#+ and industry2wide collaboration.
This section has e*amined the current situation in the 3.4. oil industry. It has
highlighted the increasing comple*ity of the business environment of petroleum
companies operating in the 3.4.. The ne*t section shows why, given these recent
changes, the industry provides such a useful e*ample in which to study investment
decision2making. The effect of these changes on the investment decision2making
?5
processes adopted by companies in the industry will then be e*amined using the
results from recent studies of current practice.
5(7 I-?)0,M)-, A''RAI0A+ D)*I0IO--MA.I-G I- ,) OI+ I-DU0,RC
$isk and uncertainty are inherent in petroleum e*ploration &:ailey et al., in pressF
Simpson et al., 9>>> and 5666F -amb et al, 5666F !atson, 566=F /ewendorp, 566@F
$ose, 56=7F Ikoku, 56=<F +egill, 5675 and 5676'. The circumstances that led to the
generation of oil and gas are understood only in a very general sense &/ewendorp,
566@F Ikoku, 56=<'. The e*istence, or more particularly the location of traps, cannot
be predicted with certainty. Aven when a trap is successfully drilled, it may prove
barren for no immediately discernible reason &Ikoku, 56=<'. Indeed, worldwide
appro*imately nine out of ten wildcat wells &which cost appro*imately R5? million to
drill offshore' fail to find commercial quantities of hydrocarbons &!atson, 566=F (ike
and /eale, 5667F Dyne 566?'. !hilst in the /orth Sea, of the 5?>,>>> square miles
of the 3.4. area that have been offered for licence, it has been estimated that only 9E
has hydrocarbons beneath it &Simpson et al., 5666'. "urthermore, the economic
factors that ultimately affect the e*ploitation of the resources are sub8ect to capricious
shifts that, it has been claimed, defy logical prediction &Ikoku, 56=<'F an effect that is
e*acerbated in the oil industry since e*ploration pro8ects require a large initial capital
investment without the prospect of revenues for ten to fifteen years &Simpson et al.,
5666'. Such observations led /ewendorp &566@' to conclude that risk and uncertainty
are frequently the most critical factors in decisions to invest capital in e*ploration. In
reality he argues, each time the decision2maker decides to drill a well, he is playing a
game of chance in which he has no assurance that he will win &/ewendorp, 566@'.
(reviously, when most e*ploration wells were shallow and drilling anomalies were
numerous and easy to locate, the upstream decision2maker was content to utilise
intuition, 8udgement, hunches and e*perience to determine which prospects to drill
&/ewendorp, 566@'. Dowever, as noted in sections 9.9 and 9.;, the worldwide
petroleum industry has changed, and many decision2makers are uncomfortable basing
their investment decisions on such an informal approach &#hapter @F :all and Savage,
5666F /ewendorp, 566@'. #onsequently, decision analysis tools, which allow risk and
uncertainty to be quantified, have recently begun to receive increasing attention in the
?9
oil industry investment appraisal literature &for e*ample, :ailey et al., in pressF
Simpson et al., 9>>> and 5666F -amb et al., 5666F )alli et al., 5666F :all and Savage,
5666F Schuyler, 5667F +urtha, 5667F Smith and +c#ardle, 5667F %tis and
Schneiderman, 5667F /angea and Dunt, 5667F /ewendorp, 566@'.
There have been two recent studies into current practice in investment appraisal
across the oil industry &Schuyler, 5667F "letcher and .romgoole, 566@' and both have
sampling limitations. SchuylerBs study drew only on the observations from
participants on a decision analysis course and so some formal bias to formal decision
analysis practices must be assumed and "letcher and .romgoole only included sub2
surface employees in their cross2company survey. Dence, the results from both pieces
of research can only be regarded as indicative rather than conclusive. :oth studies
tend to suggest that decision analysis is receiving increasing attention in the industry
and that the techniques are widespread in e*ploration investment decision2making but
have yet to be applied to production investment decisions &the differences in these
types of investment decision will be discussed in #hapter ?'. The most useful
indication of current practice in investment appraisal comes from individual
companies publishing details of their approach to investment appraisal in the oil
industry literature. 3nfortunately there are few such reports and those that there are
usually tend to describe how decision analysis has been used in specific cases. This in
itself is indicative of how organisations use the techniques, a point that receives
further attention in #hapter @, but, moreover, make it impossible to use such accounts
to describe company2wide practice. There are a few e*ceptions. "or e*ample, %tis
and Schneidermann &5667' describe the decision analysis approach used in #hevron
and /angea and Dunt &5667' outline that used in +obil prior to their merger with
A**on. These and the other similar publications are reviewed in #hapters ? and @.
#learly though, these company accounts, by their nature, are not indicative of industry
practice.
This section has shown why, given the recent changes in the operating environment of
the industry that it provides such a useful e*ample in which to study investment
decision2making. The effect of these changes on investment decision2making in the
industry has been e*amined using the results from recent studies of current practice.
This has highlighted the growing interest in decision analysis in the industry and
?;
identified the need for an empirical study to investigate investment appraisal decision2
making in the oil and gas industry.
5(8 *O-*+U0IO-
This chapter has used the oil industry literature to present a brief description of the
industry. The main challenges facing the industry in the 95
st
century were identified.
The effects of these global changes on the 3.4. industry were e*amined. This
highlighted the growing comple*ity of the business environment of those companies
operating in the upstream oil and gas industry that has prompted increasing interest in
decision analysis techniques in the industry &#hapter @'. The chapter showed how
there have been limitations in the recent studies into current practice in investment
appraisal in the oil industry and that therefore there is a need for a study to investigate
investment decision2making in the oil and gas industry. The following chapter first
states the methodological approach adopted for this study and second, evaluates its
effectiveness.
?<
*hapter 7
Methodology
??
7(3 I-,RODU*,IO-
The current study contributes to the current deepening understanding of the value of
the application of decision analysis to organisational investment decision2making. Set
in the conte*t of the operating companies in the 3.4. oil industry the research has
three specific ob8ectives. It aims firstly to propose which decision analysis techniques
are the most appropriate for upstream oil and gas companies to utilise in their
investment decision2makingF secondly, to ascertain which of these tools upstream
companies choose to use in their investment appraisal and whyF and lastly, to establish
if there is a relationship between the use of decision analysis in investment decision2
making and good organisational performance in the operating companies in the
upstream oil and gas industry.
A qualitative methodology was chosen to answer these research questions with semi2
structured interviews being chosen as the primary research method. The interview
transcripts allowed the researcher to be able to model companiesB investment
decision2making processes and, in particular, organisationsB use of decision analysis
techniques in investment appraisal. Then, using this model, together with published
financial measures and other criteria indicative of organisational performance in the
upstream, non2parametric statistical analysis was employed for the e*amination of the
relationship between the use of decision analysis in investment appraisal decision2
making and organisational performance. In this chapter, this methodology will be
evaluated. (articular attention will be focussed on the different methods of qualitative
data analysis used and their appropriateness for the study of investment decision2
making. The choice of the oil industry as the conte*t for the current study has already
been 8ustified in the preceding chapter hence it will be taken as given here. .irections
for future research will not be discussed in this chapter but instead will be proposed in
#hapter =.
The chapter follows the approach outlined by %B+ahoney &566=' and is written as a
case history of the methodology of the current study. The chapter aims to recreate the
iterative and dynamic flows between research area and methodology that has been the
feature of several recent works. A feature of the research has been the development of
?@
the researcher as an academic researcher. In this regard, the papers and presentations
that have been prepared during the course of the current study are shown in Appendi*
9.
7(% ADO',I-G A- A''RO'RIA,) M),ODO+OGI*A+ FRAM)/OR.
%rtonBs &5667' summary of .aftBs &56=?' distinction between deductive research
&theory, method, data, findings' and inductive research &method, data, findings,
theory' suggests that the management research process can be viewed as a coherent
series of logically directed steps. )ill and ,ohnson &5665' believe that such
statements and the neat, tidy accounts of the conduct of the research process produced
by seasoned researchers &:urgess, 56=<a', are misleading. In particular, the authors
argue that they simplify concepts which are frighteningly out of step with those
researchers who have given a Hwarts and allK account of their methodologies
&%B+ahoney, 566=F :arley, 566?F :ryman and :urgess, 566<F "erner, 56=6'.
.iscussions by such researchers have revealed that social research is not a set of neat
procedures. $ather, it is a social process whereby interaction between researchers and
researched will directly influence the course of action which a research pro8ect takes
&%B+ahoney, 566=F %kley, 566<F :urgess, 56=<bF Shaffir et al., 56=>F Shipman,
567@F :ell and /ewby, 5677F :ell and Ancell, 567=F Dammond, 56@<'. The research
process, and hence the methodology employed, is not a clear cut sequence of
procedures following a neat pattern, but a messy interaction between the conceptual
and empirical world, with deduction and induction occurring at the same time
&:echofer, 567< p7;'. -aing &5667' argues that the methodological framework cannot
be seen as a rigid, purely ob8ective construct. $ather, it should be perceived as a
framework, the final version of which is determined by environmental pressures. It is
within such a conte*t that the methodological framework employed in this research
has evolved. There is widespread recognition that there can be a significant gap
between the methodological approach and intentions articulated at the commencement
of the research pro8ect and that ultimately implemented &for e*ample, %B+ahoney,
566=F -aing, 5667'. #onsequently, in seeking to demonstrate the validity and
reliability of the data gathered and the results presented, it is necessary to e*amine and
evaluate critically the actual research process undertaken &-aing, 5667' and this is the
aim of this chapter.
?7
In describing the core elements of management research, )ill and ,ohnson &5665
p5?<' stress the centrality of a comprehensive review of the e*isting literature to the
research process. They describe the literature review phase of research as
constituting:
Ha critical review which demonstrates some awareness of the current state of
knowledge on the sub8ect, its limitations and how the proposed research aims
to add to what is known.K
A comprehensive review and critical appraisal of the relevant literature is thus crucial
to formulating the underlying research questions to be e*amined by the study and in
the subsequent development of the specific research instruments to be utilised in the
data gathering process. "ollowing the approach used by -aing &5667', at the outset of
this research, the literature review involved the systematic searching of a number of
ma8or databases against a list of key words and phrases. This allowed the researcher
to identify as fully as possible all published material that broadly related to aspects of
the research sub8ect. "rom this comprehensive search, relevant articles and te*ts were
obtained, analysed, annotated and classified. Subsequently, the references and
bibliographies of key articles and te*ts identified from these databases were searched
in order to follow up additional potentially relevant material. This literature review
was continually updated throughout the duration of the research process as additional
relevant material was published. The new publications, though not impacting on the
development of the underlying research questions or the specific research instruments,
enhanced the subsequent analysis of the primary data gathered during the field
research.
In seeking to e*plore the investment decision2making process of the upstream oil and
gas industry, the literature review for the study, presented in #hapters 9 and ;,
e*amined research from two different areas. "irstly, it investigated the academic
literature on investment decision2making and, in particular, that relating to decision
theory and secondly, it e*plored the literature relating to the industry and its
investment decision2making process. $eviewing these literatures highlighted gaps in
e*isting knowledge and the identification of the research questions for the current
study. These three questions are:
?=
5. !hich techniques are the most appropriate for companies to utilise in their
investment decision2makingI
9. !hich techniques do companies use to make investment decisions and how are
they used in the investment decision2making processI
;. Is there a relationship between using decision analysis techniques in investment
appraisal decision2making and good organisational performanceI
This section will e*amine the specific research instruments used to e*plore these
questions in turn. The following section will evaluate the effectiveness of the
methodological approach.
To answer the first research question and identify the decision analysis tools that are
most appropriate for investment appraisal decision2making in the upstream oil and gas
literature, the current study drew primarily on the decision theory and oil industry
literatures. This involved firstly, identifying the whole range of techniques that are
available and, secondly deciding which of these tools are the most appropriate for
upstream investment decision2making. It demanded careful consideration of factors
such as the business environment of the upstream industry and the level and type of
information used for investment decision2making in the industry. Through this
process, the research identified the decision analysis techniques that are particularly
useful for upstream investment decision2making. Then, drawing again on the
investment appraisal and industry literatures, and also on insights gained at
conferences and seminars, an approach to investment decision2making in the oil
industry was developed that utilised the full spectrum of tools identified. Some
decision analysts advocate using one decision analysis technique for investment
appraisal &for e*ample, Dammond, 56@@'. Dowever, in reality, each tool has
limitations &-efley and +organ, 5666' some that are inherent, others which are caused
by a lack of information or specification in the literature &see #hapter ?'. As such, the
knowledge that the decision2maker can gain from the output of one tool is limited &see
#hapter ? and /ewendorp, 566@'. Therefore, a combination of decision analysis
techniques and concepts should be used to allow the decision2maker to gain ma*imum
insight, encouraging more informed investment decision2making &this is 8ustified in
#hapter ?'. Some oil industry analysts have recognised this and presented the
?6
collection of decision analysis tools that they believe constitute those that decision2
makers ought to use for investment decision2making in the oil and gas industry &for
e*ample, /ewendorp, 566@'. Dowever new techniques have only recently been
applied to the industry &for e*ample, )alli et al., 5666F .i*it and (indyck, 566= and
566<F $oss, 5667F Smith and +c#ardle, 5667' and as such, these previously presented
approaches now require modification. #onsequently, although informed through
secondary data sources, the identification of the decision analysis techniques that are
most appropriate for investment appraisal decision2making and the approach to
investment appraisal that is presented in this thesis, are believed to be two of the main
findings of the research.
In e*ploring the second research question, the current study aimed to establish current
practice in investment appraisal decision2making in the operating companies in the
3.4. oil and gas industry. Two factors directly affected the choice of research method
chosen to investigate this questionF firstly, there is widespread recognition in social
science research that the primary strength of qualitative research is that it facilitates
the in2depth e*ploration of the perceptions and values of key organisational
stakeholders. :ryman &56=6 p59' identified the principal advantage of qualitative
research as being that:
Hit e*presses commitment to viewing events, actions, norms and values
from the perspective of the people being studied.K
This reflects the primarily interpretive approach inherent in qualitative research
involving the e*ploration of meanings and perceptions within a naturalistic rather than
positivist framework &Dammersley and Atkinson, 56=;'. Such essentially intangible
issues, -aing &5667' argues, cannot be e*plored adequately by traditional quantitative
survey2based research methods. A second factor that impinged on the choice of
research method was that previous empirical research had typically used quantitative
survey2based research which had produced statistical results that indicated the
percentage of organisations using decision analysis techniques &defined in #hapter ?'
&for e*ample see studies by Arnold and Dat1opoulous, 5666F #arr and Tomkins, 566=F
Schuyler, 5667F :uckley et al., 566@F Shao and Shao, 566;F 4im, "arragher and
#rick, 56=<F Stanley and :lock, 56=;F !icks 4elly and (hilippatos, 56=9F :avishi,
56=5F %blak and Delm, 56=> and Stonehill and /athanson, 56@='. $esearchers such
@>
as #lemen &5666' perceived that through using survey2based research methods these
studies had overlooked many interesting issues. "or e*ample, they had not indicated
why decision analysis was used in some organisations but not others nor had they
provided an e*planation of why companies endorsed the use of certain techniques and
failed to implement others &#lemen, 5666'. It can be argued that the failure of such
studies to e*amine these issues had contributed to the division between behavioural
decision theorists and decision analysts outlined in Section 9.< of #hapter 9.
Therefore, "ualitati,e methods were chosen to build a picture of decision analysis use
in upstream organisations. In using such an approach the researcher aimed to
investigate the relationship between normative decision analysis models and
behavioural decision theory descriptions.
The ne*t step involved deciding which companies would comprise the population for
the current study. -imited resources had already dictated that the research had to be
constrained to oil companies who had offices in the 3.4.. A list of all the operators
active in the 3.4. was obtained via 34%%A &3nited 4ingdom %ffshore %perators
Association'. These thirty2one companies then constituted the population for the
research study.
"ollowing this, the ne*t decision was which qualitative instrument, or combination of
instruments, to use. +ethods such as case study research and participant observation
were clearly impractical with such a large number of companies within a relatively
short and finite time2period. Acknowledging that the companies in the sample were
all competitors, and that the issues that would be under discussion are commercially
sensitive multi2company focus group discussions were also omitted from
consideration. Intra2company focus groups were considered and re8ected since it was
felt that this would prohibit a frank e*pression of attitudes and e*periences and
become an e*ercise in reporting the formal organisational perspective. This process
of elimination identified a semi2structured style of interviewing with several
respondents in each company to be a useful instrument to employ in this research.
3sing this technique, a set of themes and topics is defined to form questions in the
course of conversation. This strategy, it is argued in the qualitative methods literature,
gives informants an opportunity to develop their answers and allows the researcher
the freedom to follow up ideas, probe responses and investigate motives and feelings
@5
&for e*ample, %kley, 566<F :ryman, 56=6'. It also allows the researcher to be
sensitive to the participantsB understanding of the language and concepts under
investigation. In e*ploring the investment appraisal decision process, the interviews
focussed on four core areas. These were firstly, decision analysis techniques that were
used and whyF secondly, how the tools were usedF thirdly, the integration of the results
from the techniques into the whole investment decision2making processF and fourthly,
respondentsB perceptions of the effectiveness of the process. :ased on such core
themes, a common interview schedule &Appendi* 5' was developed. The initial
questions focused on relatively broad conceptual issues, progressing to specific
practical issues during the course of the interview, with the aim of producing rich and
detailed accounts of the participantsB perceptions of the investment decision2making
process.
:( volunteered to pilot the interview schedule and the researcher visited them several
times in April 566= to conduct interviews. The researcher was unfamiliar with
interviewing. Taping the interviews permitted reflection later on the researcherBs
ability and role as an interviewer. The interviews were transcribed and these
transcripts played a valuable role indicating those terms and processes the researcher
was using which were academically understood but which required clarification at
practitioner level. The interview schedule was amended accordingly with some
questions being rephrased, others discarded and several new questions being added.
These interviews were used to improve the researcherBs technique as an interviewer
rather than for data collection. After the modification of the interview schedule, the
transcripts from these interviews were discarded.
The researcher then had to decide whom to approach in each of the companies and
how best to approach them. Ideally, the researcher wanted to speak to individuals
who were actively involved in the whole investment appraisal process. !ith this
rationale, it was decided to approach each operatorBs A*ploration +anager, or
equivalent, using the membership list of the (etroleum A*ploration Society of )reat
:ritain. Initially the pro8ect was outlined in a letter that indicated what would be
required of each participant company, detailing what would be done with the collected
data and giving assurances of confidentiality and anonymity. This was then followed
up with either an e2mail or telephone call.
@9
The letters were sent in +arch 566= and the researcher began receiving responses in
early April. The number of positive responses was overwhelming. Twenty2seven of
the thirty2one companies approached agreed to participate J a response rate of =7E.
This high response rate is clearly a function of timing and sub8ect. As indicated in
#hapter ;, the increasingly dynamic and comple* operating environment of upstream
had increased the pressure on petroleum companies to manage their investment
decision2making processes better and decision analysis techniques were beginning to
receive increasing attention in the industry literature &for e*ample, :all and Savage,
5666F )alli et al., 5666F !atson, 566=F Schuyler, 5667F +urtha, 5667F /angea and
Dunt, 5667F %tis and Schneiderman, 5667F /ewendorp, 566@'. .uring the interviews,
many of the respondents reported that their organisations were currently in the process
of reviewing their investment decision2making processes and that this was their
motivation for participating in the research.
.ata collection began in +ay 566=. The interviews varied in length with the ma8ority
lasting appro*imately two hours. All the interviews were tape recorded in full. In
addition, notes were taken during the interviews to highlight key issues and facilitate
the subsequent analysis of transcripts. After assurances of confidentiality and
anonymity, none of the respondents had any reservations about such recording. This
emphasis on confidentiality inevitably influences the way in which the data is
subsequently utilised and presented. All the companies interviewed have been
assigned a code letter. The letter that was assigned to each company depended on the
number of decision analysis techniques used in their investment appraisal approach.
#ompany A used the least number of decision analysis techniques. #ompany : used
the ne*t least number of decision analysis tools and so on. These labels have been
used throughout the thesis. Although 97 companies were originally interviewed,
subsequent merger activity reduced this number to 9>. So only letters A to T have
been assigned to represent the interviewed companies, with letter T representing the
company that used the highest number of decision analysis techniques. !here more
than one respondent was interviewed in an organisation, each interviewee was
assigned a number so that, for e*ample, the second respondent that was interviewed
from company : would be referred to as :9 and the third as :;. It is important to
note that where companies merged, the respondents in these organisations were
@;
contacted after the merger and asked to report any changes to their corporate
investment appraisal process. These insights were then analysed along with the
relevant interview transcripts in the ne*t stage of the research.
!hile divisions e*ist amongst researchers over the issue of whether interviews should
be transcribed selectively or in full &:ryman, 56=6', given the emphasis within this
research on securing an in2depth understanding of attitudes and e*periences, it was
decided to transcribe all interviews in full. Such an approach, though time2
consuming, facilitated the identification of themes, utilisation of quotations and the
avoidance of biased 8udgements arising from initial impressions of the interview data.
In addition, when coding the interview data from the transcripts, the original tapes
were utilised, alongside the contemporaneous notes, in order to ensure that the
intervieweesB e*pression and emphasis was taken into account &-aing, 5667'. !here
necessary, respondents were contacted by telephone or e2mail for clarification.
The challenges of analysis and interpretation of qualitative data are widely recognised
and well documented &$ossman and $allis, 566=F :ryman and :urgess, 566<F
Dammersley, 5669F .en1in, 567='. The difficulty of handling such data is well
illustrated by +ilesB &5676' description of qualitative data as Han attractive nuisanceK.
In analysing the data from this research, rigorous use was made of appropriate
structural approaches such as inductive analysis. In inductive approaches to data
analysis, hypotheses are not generated before the data are collected and therefore the
relevant variables for data collection are not predetermined. The data are not grouped
according to predetermined categories. $ather, what becomes important to analyse
emerges from the data itself, out of a process of inductive reasoning &+aykut and
+orehouse, 566< pp59@2597'. In this research pro8ect, the analysis of the interview
data involved the coding of this data against both the core themes contained in the
interview schedule which were derived from the analysis of the relevant literature and
the emergent themes identified through the contemporary notes. After this initial
coding, the data was further coded under more specific themes as well as additional
emergent themes. Such multi2stage coding is vital in order to avoid as far as possible
constraining any potential empirically based conceptual development to flow from
this research &.en1in, 567='. It must be noted that while the data collection and data
analysis elements of the research are described separately here, they cannot be seen as
@<
discrete stages &-aing, 5667'. In common with many other qualitative studies, the
collection and inductive analysis of the data ran concurrently although the balance
between the two elements shifted over the duration of the research. %kley &566<
pp9>295' writes that
H...to the professional positivist this seems like chaos. The fieldworker
cannot separate the act of gathering the material from that of its continuing
interpretation. Ideas and hunches emerge during the encounter and are
e*plored or eventually discarded as fieldwork progresses.K
!iseman &567< p;57' writes that this constant interplay of data gathering and analysis
is the essence of qualitative research. It facilitates the flow of ideas between the two
processes and this contributes to the development of theoretical constructs
&Aisenhardt, 56=6'. +oreover, whilst the author has attempted here to detail the
techniques used to assist in the data analysis, the precise mechanism by which this
occurred cannot be fully documented. This point is echoed by %kley &566< p95':
HAfter the fieldwork the material found in notebooks, in transcripts and even
contemporary written sources, is only a guide and trigger. The anthropologist
writer draws on the totality of the e*perience, parts of which may not, cannot be
cerebrally written down at the time. It is recorded in memory, body and all the
senses. Ideas and themes have worked through the whole being throughout the
e*perience of fieldwork. They have gestated in dreams and the subconscious in
both sleep and waking hours, away from the field, at the anthropologistBs desk, in
libraries and in dialogues with people on return visits.K
!ith so much of the data analysis taking place in the sub2conscious mind, it is
impossible to present a full account of it &!hyte, 56?? p976'. The current study then
uses the approach of -aing &5667' who believes one way to ensure the integrity of the
data and the ob8ectivity of the resultant findings is for researchers to use verbatim
accounts taken within their original conte*t.
Through this process a description of current practice was produced and, therefore,
research question two was answered. As indicated earlier, through the answering the
first research question, the quantitative techniques available to companies for
investment appraisal decision2making had been identified and an approach to
investment decision2making developed that utilised the full spectrum of available
techniques. These both acted as input into the third stage of the research which aimed
@?
to establish if there is a relationship between the use of decision analysis in
investment appraisal decision2making in companies and business success in the
operating companies in the 3.4. upstream oil and gas industry.
As indicated in #hapter 9, very few studies have attempted to value the usefulness of
a decision analysis approach &#lemen, 5666'. #lemen and 4wit &9>>>' are the only
researchers who have attempted to evaluate the benefits of a decision analysis
approach to managing risk and uncertainty in investment decision2making. These
authors investigated the value of a decision analysis approach within one company
using a qualitative methodology, specifically using depth interviews and documentary
analysis to inform their research. This methodological approach permitted these
researchers to value the HsoftK benefits to an organisation of utilising a decision
analysis approach. Dowever, whilst their research provides useful insights, as the
authors themselves acknowledge, the focus on one organisation means that the results
cannot be generalised to a larger sample. The current study differs from this since it
aims to produce an indication of the value of using a decision analysis approach in
upstream oil and gas companies. Therefore, by implication, the research involves
numerous companies and this prohibits use of the time2consuming qualitative
methodology implemented by #lemen and 4wit &9>>>'. Instead, using the results
from the semi2structured interviews as input, it was assumed that any value added to
the company from using a decision analysis approach, including HsoftK benefits,
ultimately affects the bottom2line &this assumption is 8ustified in #hapter 7'. It was
then possible to investigate the relationship between an organisationBs use of decision
analysis and good decision2making statistically by using criteria that are indicative of
organisational performance.
To permit a comparison of companies according to the decision analysis techniques
used for investment appraisal &identified by answering the second research question',
a ranking scheme was devised which assigned two points for full implementation of
each of the techniques identified by answering the first research question, one point
for partial implementation of, or some familiarity with, the technique, and 1ero for
non2use. #ompanies were also graded in the same way on how well the decision
analysis techniques were supported J specifically, if there had been an attempt to
introduce corporate definitions of the key terms risk and uncertainty. :est practice
@@
companies were e*pected to have implemented definitions that were complementary
to their decision analysis approach. !here there were numerical ties according to
these criteria, the tie was broken on the basis of other material from the interviews,
indicative of level of sophistication, which was not available on all companies and
therefore not included as an overall rank measure &for e*ample, company2wide
application of a piece of software'. This ranking scheme is discussed further in
#hapter 7. (erformance measures were then selected that are indicative of business
success in the upstream. The choice of these outlined and 8ustified in #hapter 7. The
appropriate performance data was then gathered on each company. "or some of the
criteria, it was only possible to access ordinal level data. "or the some however,
categorical data was available. The relationship between the rank each company
achieved in the decision analysis ranking and their rank, or otherwise, on each
performance measure were then analysed together statistically.
There is a large number of statistical techniques available for analysing any given set
of data. The author has chosen in this thesis to use those tools known as non2
parametric or distribution2free. These techniques may be contrasted with others
known as parametric techniques. (arametric techniques make a large number of
assumptions regarding the nature of the underlying population distribution that are
frequently untestable. -each &5676' argues that social scientists using parametric
statistical analysis are taking a gamble. If the population assumptions are correct or
appro*imately correct, then the researcher has very good test. Dowever, if the
population assumptions are incorrect, then a non2parametric test may well give a more
accurate result. /on2parametric tests make relatively few assumptions about the
nature of the data and hence are more widely applicable. "inch and +c+aster &9>>>
p56' write:
Hnon2parametric techniques do not invoke such restrictions. Techniques
involved in measuring association do not require the employment of cardinal
measures redolent of interval scales. Instead the only measurement
requirement is that ordinal scales can be deployed &-ehmann and .BAbrera,
567?F Siegel and #astellan, 56=='.K
Since only ordinal level data were available for some of the performance criteria and
there were no ties in the data, the primary non2parametric technique that the
@7
researcher elected to use was SpearmanBs rank correlation test. It is outlined in
Appendi* ;.
SpearmanBs rank order correlation coefficient is a modified form of the more typically
used (earsonBs correlation coefficient. It is mathematically equivalent to (earsonBs
correlation coefficient computed on ranks instead of scores. ,ust as (earsonBs
correlation coefficient is interpreted as a measure of linearity, SpearmanBs correlation
coefficient can be interpreted as a measure of monotonicity. That is, SpearmanBs
correlation coefficient is a standardised inde* of the degree to which two variables
covary in a monotonic fashion. The SpearmanBs correlation coefficient inde* can
range from J5.> to 5.>. It will attain these ma*imum values when a perfect
monotonic relationship is negative or positive, respectively. The rank order
correlation coefficient will be 1ero when there is no relationship between two
variables, or when the relationship is strong but nonmonotonic. Since SpearmanBs
rank correlation coefficient is equivalent to (earsonBs correlation coefficient computed
on scores, it has some of the same characteristics. "or e*ample, since (earsonBs
correlation coefficient is equal to the regression coefficient in the special case where
the variances of the two populations are equal it now follows that SpearmanBs
correlation coefficient is also the linear regression coefficient for two ranked
variables. As such, it indicates the amount of Hrank changeK in one population when
the other increases by one rank. It can also be shown that SpearmanBs rank order
correlation coefficient indicates the proportion of variation in one population that is
e*plained by variation in the other population. SpearmanBs rank order correlation test
is considered by some statisticians to be a Hquick and dirtyK appro*imation for
(earsonBs correlation coefficient. Dowever, when data are ordinal the (earsonBs
correlation coefficient is not appropriate. In this case, SpearmanBs correlation
coefficient is the most desirable inde* &-each, 5676'.
!here categorical data was available and there was sufficient number of data points
on a performance criterion, a 4ruskal !allis test was also used. This test is outlined
in Appendi* <. The 4ruskal !allis test is a direct generalisation of the !ilco*on
$ank Sum test. !hen a significant result is obtained with the 4ruskal !allis test, all
that can be concluded is that there is some difference in location between the samples.
@=
To find the location of this difference, the !ilco*on $ank Sum test was used. This is
also outlined in Appendi* <.
Through the utilisation of these two non2parametric tools, the author was able to
produce evidence of an association between good organisational performance and the
use of decision analysis techniques in investment appraisal decision2making in the
operating companies in the 3.4. upstream oil and gas industry.
This section has outlined the research methodology used to answer the three research
questions proposed in #hapter 5. The following section will assess its effectiveness
and suggest possible improvements.
7(5 )?A+UA,I-G ,) )FF)*,I?)-)00 OF ,) R)0)AR* M),ODO+OGC
Through the utilisation of qualitative methods and statistical analysis the research
presented in this thesis has generated a robust body of data. This claim can be
8ustified on three counts.
"irst, using the academic investment decision2making and oil industry literatures, an
approach to investment decision2making has been developed that utilises the full
complement of decision analysis techniques presented in the decision theory
literature. Second, using semi2structured interviews a description of the use of
decision analysis in investment decision2making in the upstream was produced that is
representative of current practice in the industry. The data from this second stage are
consistent with previous research that indicated a gap between current practice and
current capability in investment appraisal &for e*ample see studies by Arnold and
Dat1opoulous, 5666F #arr and Tomkins, 566=F Schuyler, 5667F :uckley et al., 566@F
Shao and Shao, 566;F 4im, "arragher and #rick, 56=<F Stanley and :lock, 56=;F
!icks 4elly and (hilippatos, 56=9F :avishi, 56=5F %blak and Delm, 56=> and
Stonehill and /athanson, 56@='. The findings as such are demonstrably valid and
reliable. Thirdly, using the data gathered from the research interviews and publicly
available financial data, the relationship between use of decision analysis techniques
in investment appraisal decision2making and good organisational performance was
investigated statistically. As such the research has contributed to the current
@6
discussion &for e*ample, #lemen and 4wit, 9>>>F #lemen, 5666' in the decision
theory literature of the relationship between normative and descriptive models of
investment decision2making. "urthermore, since the sample used in this research
contains =7E of the 3.4. petroleum operators, it is clear that the findings are
representative of all 3.4. petroleum operators. +oreover, since most of the oil
companies that operate in the 3.4. are amongst the ma8or players in the oil industry,
the findings can be said to be indicative of investment decision2making practices in
the ma8or companies in the worldsB upstream oil and gas industry.
As with any research, the results have to be interpreted bearing in mind some
limitations. The conte*t within which the research is undertaken inevitably impinges
on the actual articulation of the research methods employed. In this regard the time,
and by implication the resource limitations, influenced the final methodology adopted
in three ways.
"irstly, time and resource constraints precluded the use of observational research
techniques that would have facilitated an enhanced understanding of the dynamics of
the investment decision2making process and the links between the different stages of
the process, the HsoftK effects on organisation performance from using decision
analysis and the relationships between the individuals involved.
Secondly, the research was limited to a single time period that coincided with a period
of very low oil prices, proliferation of mergers and corresponding 8ob losses.
Interviewing respondents who knew that they were to be made redundant affected the
data gathered from them since respondents often perceived their organisationsB
approach to decision2making as e*tremely poor and portrayed their management in a
less than favourable light. In all cases this data was disregarded and other respondents
were used in these companies. "urthermore, there was tremendous uncertainty in the
industry at this time and many companies were changing their approach to investment
decision2making and, as indicated above, were becoming more interested in decision
analysis. %ften the respondents from these companies were actively involved in this
change and, on many occasions, they perceived the current study as a vehicle for
initiating or encouraging it. As indicated in section <.9, this significantly affected the
response rate and also resulted in these respondents being particularly forthcoming
7>
with information on company practice. If it had been possible to return to these
companies, it would have been interesting to look further on the effect of the mergers
on companiesB investment appraisal decision2making practices.
Thirdly, resource and time constraints affected the number of people consulted in each
company. .espite the initial intention to consult multiple respondents in each
company this was often not possible. Typically, only one person in each company
was interviewed usually this individual was the A*ploration, #ommercial or -icense
+anager. :owman and Ambrosini &5667' have described the dangers of using single2
respondents to ascertain company practice. They argue that where single2respondents
are used, the researchers must be convinced that the research results are not dependent
on the person that happened to be surveyed or interviewed. In this study, where the
researcher felt there was the possibility of bias from a particular company
representative, this data was disregarded and another respondent in that organisation
sought. Dence, the researcher is confident that the data gathered is representative of
company practice. "urthermore, in this study, as indicated above, it was the
A*ploration +anagers that were usually interviewed. In most companies A*ploration
+anagers are usually involved in generating investment proposals and presenting
these proposals to the boards of their companies. So, in most cases, this individual is
often the only person in the company who has been involved in both generating the
analysis and witnessing the investment decision2making process.
7(7 *O-*+U0IO-
The methodology that was used to e*plore the research questions set out in #hapter 5
has been described and critically evaluated in this chapter. In doing so, it has
provided an e*ample of how research can differ from the ordered and rational
approaches of the more prescriptive research methodology te*ts. The limitations of
the current study have been highlighted. .irections for future research will be
proposed in #hapter =.
The following chapter presents the results from the first stage of the research.
Specifically, it aims to identify the tools that are available to the upstream oil and gas
industry for investment appraisal decision2making.
75
*hapter 8
*urrent capa#ility in investment appraisal in
the upstream oil and gas industry
79
8(3 I-,RODU*,IO-
This chapter focuses on answering the first research question proposed in #hapter 5.
It draws on the discussion in #hapters 9 and ;, the decision theory and industry
literatures and insights gained at conferences and workshops, to present the range of
decision analysis techniques that are applicable to the upstream oil and gas industry
for investment appraisal decision2making. This constitutes current capability in
investment appraisal in the upstream. It will be compared with current practice in
#hapter @ and used in #hapter 7 to construct a ranking of companies according to the
sophistication of their investment appraisal approach.
There are numerous tools described in the decision theory and industry literatures for
investment appraisal decision2making. All the techniques presented in this chapter
allow risk and uncertainty to be quantified and have been applied to upstream
investment decisions in the literature. Aach of the techniques has limitations, so that
reliance by decision2makers only on the output from one tool for investment decision2
making is inadvisable. :y combining the output from a variety of tools, the decision2
maker is more likely to assess the risk and uncertainty accurately. The techniques
described in this chapter all use similar input and, hence their use together does not
place unnecessary strain on the resources of an organisation. The other tools
described in the literature &for e*ample, the analytic hierarchy process &Saaty, 56=>'
and +arkov chain analysis &.as et al.O5666'' have either not been applied to the
upstream in the literature, or the input they demand and, in many cases, the output
they produce, is not complementary to the other investment techniques currently used
by organisations. Dence their use would represent a significant amount of additional
work for organisations. "or these reasons, the tools described in this chapter, the
researcher believes, comprise the toolkit currently available to the upstream
investment decision2maker.
The chapter begins by presenting the decision analysis techniques described in the
industry and prescriptive decision theory literatures. "undamental to decision analysis
are the concepts of e*pected monetary value &A+0' and decision tree analysis. They
are widely applied to investment decision2making in the oil industry &see #hapter @
7;
and studies by Schuyler &5667', "letcher and .romgoole &566@' and /ewendorp
&566@''. The chapter proceeds to e*plore the other techniques in a similar manner.
"ollowing this there is a discussion of how these techniques could be used by the
upstream for investment appraisal. This indication of current capability is then used
as input into #hapter @, where current theory in investment appraisal in the upstream
is compared with current practice. &It is important to note that in this chapter
descriptions of standard techniques are referred to, rather than reproduced. !here
necessary, the reader is referred to the relevant literature for further details.'
8(% ,) *O-*)',0 OF )I')*,)D MO-),ARC ?A+U) A-D D)*I0IO- ,R)) A-A+C0I0
A general understanding of the oil industry can be gained from the process model
shown in figure ?.5. In particular it highlights the e*tent to which investment
appraisal decision2making in the upstream is characterised by risk and uncertainty as
indicated in #hapter ;. The figure indicates the points at which investment decisions
are taken to proceed with or to abort the pro8ect. If the decision is to continue, a
further investment decision must be taken on whether to invest in the gathering and
analysis of additional data in order to assess better the risk and uncertainty &at
abandonment, the decision is not whether to abandon but when and how to do so'. At
any of these decision points, the consequences of that investment decision on all the
subsequent processes, right through to the abandonment phase, need to be estimated
and considered in the investment decision2making. "or e*ample, when a company is
considering drilling a further appraisal well in a field, an estimate of the total
recoverable reserves from the field needs to be produced and used as input to the
economic analysis &-ohren1 and .ickens, 566;'. The economic analysis will then
model the cash flow throughout the pro8ectBs life including a prediction of when the
field will be abandoned and the estimated cost &Simpson et al., 9>>>'.
The upstream oil and gas industry shares with some other businesses, such as the
pharmaceutical industry and aerospace engineering, typically long payback periods.
(ayback is defined as the length of time between the initial investment in a pro8ect by
the company and the generation of accumulated net revenues equal to the initial
investment &figure ?.9'. In the oil industry, this period is typically between ten to
fifteen years.
7<
"igure ?.5: )he upstream oil and !as industry: a multi-sta!e decision process
"or e*ample, in the /orth Sea there is an average gap of seven years between initial
e*ploration e*penditure and the commitment to develop a prospect. It takes another
three or four years to get to the point when oil is actually produced and then fields
normally produce for around twenty years before they are abandoned. &It should be
noted that currently average lead times are being reduced through the wider
availability of infrastructure and technology'. +ost of the main costs or cash
(roduction
Operations Decisions
Abandonment
,iming decision
)ravity
+agnetics $egional
geology
Dydrocarbon
(otential
$isked
economic
value
A+0
(rospect
definition
and map
.rillNno drill
J3&M decision
Astimated
reservoir
properties
Seismic
structural geology
$isk 3ncertainty
y
Sequence
stratigraphy
(roduction
rates
.evelopment
J8&&M decision
$eservoir engineering
.evelopment engineering
Aconomics
Aconomics
Aconomics
"acilities
-og analysis
!ell maintenance N
intervention
(ressure
maintenance
3llage Infill drilling
<.
seismic
Aconomics
Aconomics
Anvironment
Appraisal
drilling
J%M decision
I
I
I
I when
I
I
I
I
7?
outflows are incurred in the earlier, e*ploration and development, years while the cash
inflows or revenues are spread over the active productive lifetime of the field. This
makes economic modelling particularly difficult since at each investment decision
point indicated in the process map in figure ?.5, estimates must be generated of the
values in a decadeBs time of variables, some of which are notoriously volatile, such as,
oil price and inflation. It also means that it is critical that discounted cash flow &.#"'
techniques are adopted in investment appraisal &Simpson et al., 5666'. The most well
known .#" tool is the net present value &/(0' method and it will be reviewed here.
The intention is to give only a brief overview of /(0. +ore detailed e*planations can
be found in finance and economics te*ts &for e*ample, Atrill, 9>>>F :realey and
+yers, 566@F .rury, 56=?F !eston and :righman, 567='.

*umulative
net cash
positions
>
(ayback ,ime
period

Initial investment
"igure ?.9: Cumulati,e cash position cur,e (source: adapted from @e+endorp, 1998 p1B
As indicated above, when money is invested in a pro8ect a commitment of funds is
generally required immediately. Dowever, the flow of funds earned by the investment
will occur at various points of time in the future. #learly, receiving S5>>> in, for
e*ample, a yearBs time is less attractive than receiving S5>>> now. The S5>>> now
could be invested so that in a yearBs time it will have earned interest. This implies that
money that will be earned in the future should be discounted so that its value can be
compared with sums of money being held now. This process is referred to as
discounting to present value &)oodwin and !right, 5665 p5<7'.
Total net profit from
investment
7@
The severity with which future sums of money are discounted to their present value is
a function of the discount rate applied. .etermining the appropriate discount rate for
a companyBs potential investment pro8ect is, ultimately, a matter of 8udgement and
preference. Dowever, many attempts have been made to make the choice of a
discount rate as Hob8ectiveK as possible, making this a comple* area which is beyond
the scope of this thesis. Adinburgh based oil industry analysts !ood +acken1ieBs
base case nominal discount rate is made up of four different elements:
The risk2free real rate of return available through an inde*2linked, long2term gilt
yield. This comprises the real rate of interest known at the time of purchase and
whatever inflation rate occurs over the period of redemption.
An assumption of the long2term inflation rate.
The equity risk premium. This is the return e*pected by equity investors over and
above the return on risk free assets. A premium is required because equity returns
J like upstream investments J can only be estimated and are not guaranteed.
The e*ploration risk premium. %il companies are generally perceived as being
HriskierK than the equity market &!ood +acken1ie, 566='.
"or many situations it is convenient to let the discount rate reflect the opportunity cost
of the capital which is being invested. +ost firms now using the /(0 measure of
profitability appear to be using discount rates in the range of 6E to 5?E for petroleum
e*ploration investments. Some companies adopt a higher discount rate as a crude
mechanism for quantifying risk and uncertainty &/ewendorp, 566@ pp;?2;@'. This is
a practice that is not encouraged by many theorists since it does not e*plicitly
consider the varying levels of risk between competing investment options &for a full
discussion see /ewendorp, 566@ pp;>72;>='.
Daving determined the discount rate, the process of discounting future sums of money
is very straightforward. If the discount rate is equal to i
/(0,
the net cash flow of year k
is equal to #"
k
and the pro8ect life is equal to n years, the /(0 is given by &$oss,
5667 p<>':
n
/(0 T #"
i
L5N&5Ui
npv
'
i
M
77
iT5
If the /(0 is positive, the required rate of return will be earned and the pro8ect should
be considered &the si1e of the /(0 is often used to choose between pro8ects that all
have a positive /(0'. If /(0 is negative, the pro8ect should be re8ected.
Table ?.5 provides an e*ample of .#" analysis. It shows that at a 5>E discount rate,
the value of a R9>>> net cash flow &R9?>> of revenues less R?>> of operating
e*penses' that is received in year ? is worth R59<9 now. If R?>>> is invested today
the total /(0 &that is the sum of all the discounted net cash flows' is R9?=9. In other
wordsF the R?>>> is recovered, plus a 5>E return, plus R9?=9. If the R?>>> had been
invested in a bank at 5>E interest, an investor would have been R9?=9 worse off than
he would have been by investing in this pro8ect &:ailey et al., in press'.
C)AR I-?)0,M)-, R)?)-U) O')RA,I-G
)I')-DI,UR)
-),
*A0
F+O/
3&K
DI0*OU-,)D
*A0 F+O/
%&K
DI0*OU-,)D
*A0 F+O/
>
5
9
;
<
?
R?>>>
R9?>>
R9?>>
R9?>>
R9?>>
R9?>>
R?>>
R?>>
R?>>
R?>>
R?>>
R2?>>>
R9>>>
R9>>>
R9>>>
R9>>>
R9>>>
R2?>>>
R5=5=
R5@?;
R5?>;
R5;@@
R59<9
R2?>>>
R5@@7
R5;=6
R55?7
R6@?
R=><
,O,A+
F8&&& F3%,8&& F%8&& F8&&& F%8;% F6;%
Table ?.5: :iscounted cash flo+ concept (source: .ailey et al., in press
+ost of the companies that use /(0 as their principal Hno riskK profit indicator in
investment appraisal decision2making do so in con8unction with a sensitivity analysis
&/ewendorp, 566@'. %nce they have generated the /(0 for a particular investment
pro8ect, sensitivity analysis is used as a mechanism for investigating whether the
decision to invest would change as the assumptions underlying the analysis are varied.
Sensitivity analysis can involve varying one, two, or all the parametersB values
simultaneously &/ewendorp, 566@'. Spider diagrams are commonly used to present
the results of a sensitivity analysis with the sensitivity of the /(0 to each factor,
reflected by the slope of the sensitivity line &figure ?.;'. As the curve for a variable
becomes steeper, then changes in this parameter will result in large changes of the
dependent variable. As the curve becomes flatter, the implication is that changes in
7=
the value of the parameter cause very little change in the dependent variable
&/ewendorp, 566@ pp@@>2@@9'.
Sensitivity analysis is simple to use and it allows the analyst to focus on particular
estimates. Dowever, it does not evaluate risk and interrelated variables are often
analysed in isolation giving misleading results &Atrill, 9>>> p5@?'.
!hile /(0 is widely used, it has several disadvantages. %ne such disadvantage is
highlighted here. The others are discussed in the section of this chapter that is
concerned with option theory &section ?.@'.

* 1
D
A
-'?
/(0 base case

"igure ?.;: )ypical spider dia!ram (+here @%C is the dependent ,aria'le and #, ., C and : are
factors in the economic analysis
The /(0 approach assumes the values of the input parameters are known. "or
e*ample, in the case of the petroleum industry, its use presumes the analyst knows the
original oil2in2place, decline rate, the oil price for each year of production, costs for
each year, discount rate and ta* structure, amongst others &)alli et al., 5666'.
Dowever, in almost all cases, there is uncertainty surrounding the input variables.
A*pressing such parameters as single figures creates an illusion of accuracy. It also
means that the decision2maker has no indication as to how reliable the resulting
decision2making criterion is. #learly, it would be much more realistic if there was a
mechanism for incorporating the uncertainty surrounding the cash flows into the
analysis. As indicated in #hapter 9, decision analysis techniques now e*ist and these
Ratio o" varied parameter to #ase case
76
allow the dimensions of risk and uncertainty to be incorporated into investment
decision2making &/ewendorp, 566@ p?='.
"undamental to decision analysis are the concepts of A+0 and decision tree analysis.
:oth these tools have received much attention in the decision analysis literature and
have been applied to numerous real and hypothetical e*amples in the industry
literature. In this section, the two concepts will be briefly outlined. (articular
attention will be focussed on their impact on investment decision2making in the
upstream.
The A+0 concept is a method for combining profitability estimates of risk and
uncertainty to yield a risk2ad8usted decision criterion. The e*pected value decision
rule holds that when choosing among several mutually e*clusive decision alternatives,
all other factors being equal, the decision2maker should accept the decision alternative
which ma*imises the A+0. The A+0 of a decision alternative is interpreted to be the
average monetary value per decision that would be realised if the decision2maker
accepted the decision alternative over a series of repeated trials. The key words in this
interpretation, particularly for e*ploration decisions, are Hper decisionK and Hrepeated
trialsK as /ewendorp &566@ p@7' emphasises in the following e*cerpt:
HIf the decision2maker consistently selects the alternative having the highest
positive e*pected monetary value his total net gain from all decisions will be
higher than his gain realised from any alternative strategy for selecting
decisions under uncertainty. This statement is true even though each specific
decision is a different drilling prospect with different probabilities and
conditional probabilities.K
This statement, he argues, is the essential element for any rational 8ustification of the
use of e*pected value in business decisions. The remark suggests that, as long as
whenever the decision2maker makes an investment, he or she adopts the strategy of
ma*imising e*pected value, then he or she will do better in the long run than another
decision2maker would do by using any other strategy for selecting decision
alternatives under conditions of risk and uncertainty. #onsequently, /ewendorp
&566@' believes A+0 should be seen as a strategy, or philosophy, for consistent
decision2making rather than an absolute measure of value. "urthermore, the A+0
strategy can only be applied to advantage if used consistently from day to day:
=>
HThe decision2maker cannot use e*pected value today, some other criterion
tomorrow, and yet a third criterion on the third day.K &/ewendorp, 566@ p@7'
!hilst some decision2makers have re8ected A+0 since they believe it is difficult, if
not impossible, to assign the probabilities to the variables used in e*pected value
computations &/ewendorp, 566@ p6;', the concept has been gaining increasing
acceptance in investment decision2making in the upstream as the business
environment has become more comple* as outlined in #hapter ; &Schuyler, 5667F
Section @.5 of #hapter @'. Although each drilling decision is essentially unique, a
decision2maker may, over time, make a large number of investment decisions that
involve similar monetary sums so that the returns will still be ma*imised by the
consistent application of this criterion. This has led some to argue that the A+0
concept is perhaps particularly applicable to large organisations since they usually
have the resources to sustain losses on pro8ects that represent only a small part of their
operations &)oodwin and !right, 5665 p@?'. This may e*plain why some small
e*ploration companies have re8ected using A+0 &/ewendorp, 566@F Section @.9 of
#hapter @'. Dowever, it is arguable that the smaller company ought to be even more
aware of risk and uncertainty because of their smaller asset position and the
possibility of HgamblerBs ruinK from bad decision2making. Dence, the more likely the
rationale for the failure of some small companies to use A+0, is that they lack the
resources to conduct the necessary computations &Section @.; of #hapter @'.
The easiest way to illustrate how to calculate A+0 is to use a decision tree. A
decision tree is a tool that encourages the decision2maker to consider the entire
sequential course of action, before the initial decision &/ewendorp, 566@'. It is
accepted, almost universally, that decision trees provide decision2makers with a useful
tool with which to gain an understanding of the structure of the problems that confront
them. 4eeney &56=>' writes:
H%ften the comple* problems are so involved that their structure is not well
understood. A simple decision tree emphasising the problem structure which
illustrates the main alternatives, uncertainties, and consequences, can usually
be drawn up in a day. /ot only does this often help in defining the problem,
but it promotes client and colleague confidence that perhaps decision analysis
can help. It has often been my e*perience that sketching out a simple decision
=5
tree with a client in an hour can lead to big advances in the eventual solution
to a problem.K &)oodwin and !right, 5665 p55?'
!ells &56=9' believes users find decision trees a clear way of understanding the
issues.
The following e*ample illustrates both A+0 and the decision tree concepts. The
e*ample is taken from )alli et al. &5666'.
Suppose that an e*ploration well led to the discovery of a field that could have large
or small reserves. In the first case, installing a large platform is optimal, while
installing a small one is more appropriate in the second case. Installing the wrong
si1e platform is an e*pensive mistake. The engineer in charge of the pro8ect wants to
obtain more information before making a decision, but this is costly. !hat is the best
decisionI
"igure ?.< shows the decision tree corresponding to this situation.
"igure ?.< : #n e0ample of a decision tree (the fi!ure under the decision node is the ,alue of the 'ranch
(source: adapted from -alli et al., 1999
.ecisions are represented by squares sometimes referred to as act forks &for e*ample,
Dammond, 56@7'. The branches emanating from these correspond to possible
5;=
AppraisalNsmall or large
platform
-arge
platform
Small platform
Appraisal
-arge field
>.<
Small field
>.@
Small field
>.@
-arge field
>.<
-arge field
>.<
Small field >.@
57>
55>
5?>
5;>
5@?
59?
5;<
5<5
.)C
Act fork
Avent fork
)reen numbers A+0 of decision
alternative
=9
decisions &for e*ample, installing a large platform immediately, installing a small one,
or getting additional information'. #ircles represent uncertain &chance' events &in this
case, large reserves with a probability of @>E or small ones with a probability of
<>E'. These are sometimes referred to as event forks &for e*ample, Dammond,
56@7'. At the end of each branch &or terminal node', the final /(0 is marked. "or
e*ample, installing a large platform when the reserves prove to be large generates an
/(0 on 57> if carried out immediately compared to an /(0 of 5@? if additional
information is obtained.
To compare the decisions, the A+0 for each decision alternative is calculated at each
circular node &event fork'. "or the top branch, for e*ample, it is
57>V>.<U55>V>.@T5;<. :ecause the A+0s at the other two nodes are 5;= and 5<5,
respectively, the best decision is to carry out additional drilling before choosing the
si1e of the platform. &/ote, calculations are carried out from the terminal branches
and are Hfolded backK to the trunk.'
A large number of practical applications of these two concepts have been published
over the years. "or e*ample, 3liva &56=7' used the decision tree and A+0 concepts
to help the 3.S. postal service to decide on whether to continue with the nine2digit 1ip
code for business users. The analysis was designed to compare the monetary returns
that might result from the use of various types of automatic sorting equipment either
with or without the code. The author reported that the approach helped the decision2
makers:
Hto think creatively about the problem and to generate options.K. &)oodwin
and !right, 5665 p555'
+adden et al. &56=;' applied decision tree analysis to a problem faced by the
management of a coal2fired power plant in evaluating and selecting particular
emission control equipment. !inter &56=?' used the techniques in management union
bargaining. A number of researchers &for e*ample, /ewendorp, 566@F Dosseini, 56=@F
)rayson, 56@>' have applied decision tree analysis and A+0 to drilling decisions.
%ften these problems consider only two possible outcomes, namely success and
failure. Dowever, in some problems the number of possible outcomes may be very
=;
large or even infinite. #onsider, for e*ample, the possible levels of recoverable
reserves a company might achieve from drilling an e*ploration well. Such a variable
could be represented by a continuous probability distribution. This can then be
included in a decision tree by using a discrete probability distribution as an
appro*imation. A number of methods for making this type of appro*imation have
been proposed in the literature, the most commonly referred to is the A*tended2
(earson Tukey appro*imation &A(T'. This appro*imation technique, developed by
4eefer and :odily &56=;', based on earlier work by (earson and Tukey &56@?', is
acknowledged to generate good appro*imations to a wide range of continuous
probability distributions. "or an illustration see )oodwin and !right &5665 p55>'.
As 4eefer and :odily acknowledge however, the A(T appro*imation does have
limitations. "or e*ample, it is not applicable when the continuous probability
distribution has more than one peak or the continuous probability distribution is
highly skewed. .espite this, the technique is widely recognised as providing a useful
mechanism for generating an appro*imation for continuous probability distributions
&)oodwin and !right, 5665 p55>'.
The prescriptive decision analysis literature does not provide a normative technique
for eliciting the structure of a decision tree &some behavioural analysts have proposed
influence diagrams as a useful tool for eliciting the decision tree structure from the
decision2makerF see )oodwin and !right &5665 p55=' for a full e*planation'.
Structuring decision trees is therefore a ma8or problem in the application of decision
analysis to real problems and, clearly if the structure is wrong, the subsequent
computations may well be invalid. "ollowing such observations 0on !interfeldt
&56=>' notes that it is good decision analysis practice to spend much effort on
structuring and to keep an open mind about possible revisions. Dowever, problem
representation, according to )oodwin and !right &5665', is an art rather than a
science. "ischoff &56=>' argues similarly:
H$egarding the validation of particular assessment techniques we know
ne*t to nothing about eliciting the structure of problems from decision2
makers.K &)oodwin and !right, 5665 p55?'
+any decision2makers report that they feel the process of problem representation is
perhaps more important than the subsequent computations &)oodwin and !right,
=<
5665 p557'. Dumphreys &56=>' has labelled the latter the Hdirect valueK of decision
tree analysis and the former the Hindirect valueK. Some studies have illustrated that
the decision2makersB estimates, 8udgements and choices are affected by the way
knowledge is elicited. This literature was reviewed in #hapter 9. .espite this, the
value of decision tree analysis is undisputed, not least because decision tree analysis is
not e*clusively linked to the A+0 with its inherent and questionable assumptions
about the decision2makerBs attitude to money. These are now reviewed.
The e*pected value decision rule makes several assumptions. "irstly, it assumes that
the decision2maker is impartial to money. This assumption, not surprisingly, has been
widely criticised. Such criticisms are well illustrated by the St (etersburg (arado*
first described by .aniel :ernoulli in 57;= and outlined here.
A decision2maker is offered the following gamble. A fair coin is to be tossed until a
head appears for the first time. If the head appears on the first throw then the
decision2maker will be paid S9, if it appears on the second throw, S<, if it appears on
the third throw S=, and so on. The question is then, how much should the decision2
maker be prepared to pay to have the chance of participating in this gambleI
The e*pected returns on this gamble are:
S9V&>.?'US<V&>.9?'VS=&>.59?'U...etc.
which is equivalent to 5U5U5U to infinity. So the e*pected returns will be
infinitely large. %n this basis, according to the A+0 criterion, the decision2maker
should be prepared to pay a limitless sum of money to take part in the gamble.
Dowever, given that there is a ?>E chance that their return will only be S9, and an
=7.?E chance that it will be S= or less, it is unlikely that many decision2makers would
be prepared to pay the amount prescribed by the A+0 criterion.
Secondly, the A+0 criterion assumes that the decision2maker has a linear value
function for money. An increase in returns from S> to S5 million may be regarded by
the decision2maker as much more preferable than an increase from S6 million to S5>
million, yet the A+0 criterion assumes that both increases are equally desirable.
=?
Thirdly, the A+0 criterion assumes that the decision2maker is only interested in
monetary gain &)oodwin and !right, 5665 p@<'. Dowever, when a company is
deciding how best to decommission an offshore production facility, for e*ample, they
will want to consider other factors such as corporate image and environmental
concerns. All these attributes, like the monetary returns, would have some degree of
risk and uncertainty associated with them.
3sers of e*pected value theory have long recognised these shortcomings. As early as
579> academics were beginning to modify the concept to include the biases and
preferences that decision2makers associate with money into a quantitative decision
parameter. In essence these attempts were trying to capture the decision2makerBs
intangible feelings in a quantitative decision parameter which the decision2maker
could then use to guide 8udgements. This approach is typically referred to as
preference theory and the following section discusses this further. It draws on the
prescriptive decision analysis literature to outline first the mathematics underlying
preference theory and then proceeds to evaluate critically its contribution to
investment decision2making particularly in the upstream.
8(5 'R)F)R)-*) ,)ORC
+ost formal analyses of business decisions involving risk and uncertainty, for
e*ample the A+0 concept described above, assume that every individual or company
has, or ought to have, a consistent attitude toward risk and uncertainty. The
underlying assumption is that a decision2maker will want to choose the selected
course of action by Hplaying the averagesK on all options, regardless of the potential
negative consequences that might result, to choose the course of action that has the
highest e*pected value of profit. Dowever as Dammond &56@7' and Swalm &56@@'
observed, few e*ecutives adopt such an attitude toward risk and uncertainty when
making important investment decisions. $ather, decision2makers have specific
attitudes and feelings about money, which depend on the amounts of money, their
personal risk preferences, and any immediate andNor longer2term ob8ectives they may
have. As :ailey et al. &in press' argue:
=@
HIn the case of industries like oil where risk plays such an important part in the
thinking of e*ecutives, individual &or group' attitudes to risk and risk taking
can be important.K
Such attitudes and feelings about money may change from day to day and can even be
influenced by such factors as business surroundings and the overall business climate
at a given time &/ewendorp, 566@ p5;='.
Similar observations led Dammond &56@7' and others &for e*ample, )oodwin and
!right, 5665F Swalm, 56@@' to argue that since the A+0 concept does not include, in
any quantitative form, the consideration of the particular attitudes and feelings the
decision2maker associates with money, it may not provide the most representative
decision criterion. These writers perceive preference theory as offering a useful tool
to incorporate these attitudes and feelings regarding money into a quantitative
parameter.
The concepts of preference theory are based on some very fundamental, solid ideas
about decision2making that are accepted by virtually everyone who has studied the
theory &/ewendorp, 566@'. Dowever, the real world application of preference theory
is still very controversial and, some academics, and many business e*ecutives,
question its value in the investment decision2making conte*t. In the many articles and
books on investment decision2making, preference theory is some times referred to as
utility theory or utility curves. !hile the latter is used more frequently in the decision
analysis literature, it is also used to describe another sub8ect in economics. Dence, the
term Hpreference theoryK will be used here &/ewendorp, 566@ p5;7'. This section
will discuss first the principles of preference theory before reviewing its applicability
to investment decision2making in the upstream.
In 57;= the mathematician .aniel :ernoulli published an essay in which he noted a
widespread preference for risk aversion. In an often referred to article in Scientific
#merican in the 56=>s .aniel 4ahneman and Amos Tversky gave a simple e*ample
of risk aversion &4ahneman and Tversky, 56=9'. Imagine you are given a choice
between two options. The first is a sure gain of R=>, the second a more risky pro8ect
in which there is an =?E chance of winning R5>> and a 5?E chance of winning
nothing. !ith the certain outcome you are assured of R=>. !ith the riskier option
=7
your A+0 would be R=? &R5>>V>.=? plus R>V>.5?'. +ost people, say 4ahneman and
Tversky, prefer the certain gain to the gamble, despite the fact that the gamble has a
higher A+0 than the certain outcome &:ailey et al., in press'.
In 56<<, von /eumannn and +orgenstern e*panded preference theory and proposed
that the fundamental logic of rational decision2making could be described by eight
a*ioms that are paraphrased in the following statement:
H.ecision2makers are generally risk averse and dislike incurring a loss of RW
to a greater degree than they en8oy making a profit of RW. As a result, they
will tend to accept a greater risk to avoid a loss than to make a gain of the
same amount. They also derive greater pleasure from an increase in profit
from RW to RWU5 than they would from R5>W to R5>WU5
K
&:ailey et al., in
press'
They went on to show that if a decision2maker had a value system which was
described by these a*ioms, then there e*isted a function, or curve, which completely
described his attitude and feelings about money &/ewendorp, 566@ p5?9'. This curve
is known as a preference, or utility curve. An e*ample of a preference curve is shown
in figure ?.?.
Increasing pre"erence
or desira#ility
Increasing amounts o" money !or some other criterion$
"igure ?.?: # preference cur,e (source: adapted from @e+endorp, 1998 p1B7
According to risk consultant (eter $ose &56=7', a preference curve shows two things:
(leasure
(ain
==
The pleasure &utility' associated with winning is generally less than the
displeasure of losing the same amount &that is, it hurts more to lose than it feels
good to win.' (eople will take a greater chance to avoid a loss than to make a gain
of the same amount.
(eople feel more pleasure about gaining R5> going from, say, R5> to R9>, than
they do about gaining R5> going from R5?>> to R5?5>.
Theoretically at least it is possible to draw 8ust such a curve for any individual.
.ifferent shaped curves would denote different types of decision2maker. The shape of
the curve in the lower left2hand quadrant describes how the individual feels about loss
and the one in the upper right quadrant is the individualBs attitude to risk and the
levels of profit associated with risk. +any writers have categorised decision2makers
according to the shape of their preference curves. In general, these authors perceive
there to be three types of decision2maker: risk averters, average players &who would
always choose the decision alternative with the ma*imum A+0' and risk2seekers.
Aach, they believe, has a distinctive preference curve. These curves are shown in
figure ?.@. As indicated above, e*tensive studies &for e*ample, /ewendorp, 566@F
Dammond, 56@7' have provided strong evidence that the ma8ority of decision2makers
are risk averse to some degree, so concave downwards preference curves are the most
commonly observed in practice.
5.>
$isk averter
're"erence Averages player
$isk2seeker
>.>
Increasing amounts o" money
"igure ?.@: )ypical preference cur,es (source: Dammond, 1987
=6
%nce the decision2makerBs preference curve has been drawn, von /eumann and
+orgenstern showed that it could be used to solve decision problems using an
e*tension of decision tree analysis. The basic principle is if the decision2maker
wishes to make the decision consistent with his attitude toward risk, then the decision2
maker must choose that course of action that has the highest preference.
(reference theory is well illustrated by the following e*ample taken from Dammond
&56@7'. Imagine an oil company e*ecutive is facing the decision of whether to drill a
well. The decision2maker has three choices: firstly, drill immediatelyF secondly, pay
to acquire and interpret seismic data and then, depending on the result of the test,
decide whether to drill or notF or lastly, to let the option e*pire.
The seismic analysis can be performed for a fi*ed fee of R;>,>>> and the well can be
drilled for a fi*ed fee of R5>>,>>>. A large organisation has promised the decision2
maker that if this well discovers oil, it will purchase the companyBs rights to the oil for
R<>>,>>>. The geologists have estimated that there is >.?? probability that if a well is
drilled it will discover oil. .ata on the reliability of the seismic analysis indicate that
if the analysis is favourable, the probability of finding oil will increase to >.=?, but if
the analysis is unfavourable, it will fall to >.5. The geologists have computed that
there is a >.@ probability that the result will be favourable if seismic interpretation is
carried out.
"igure ?.7 shows the decision tree for this e*ample. At each terminal fork in the
decision tree, the e*pected value of the decision alternative is noted. &$ecall that this
is the weighted2average of the numbers at the end positions emanating from the fork'.
"or e*ample, the top2most terminal fork e*pected value is R;<>,>>>
&>.=?VR<>>,>>>U5?VR>'. $olling back the decision tree the decision2maker would
end up with the decision tree in figure ?.= and the decision, according to A+0, would
be to drill immediately. 3sing preference theory, the result is somewhat different.
To implement preference theory, assume that the decision2makerBs preference curve
has been ascertained. This is shown in figure ?.6. Then:
6>
5. #onvert all of the end positions of the decision tree into preferences &ascertained
from the decision2makerBs preference curve in figure ?.6'. These numbers are red
in figure ?.5>.
9. "ind the decision2makerBs preference for an event fork by taking the mathematical
e*pectation of the preferences values at the end position of the fork. In other
words, instead of multiplying the dollar values by probabilities, as in a decision
tree analysis using e*pected values, multiply the preferences by the probabilities.
So, at each event fork take a weighted2average of the preferences, where the
weights are the probabilities. "or e*ample, at the uppermost event fork
representing Hoil2no oilK, the preference is >.=; &>.=?V>.6;U5?V>'. The
preference is written under the fork in green in figure ?.5>.
;. "or each act fork, the decision2maker or analyst then selects the act with the
highest preference. "or e*ample, the upper most decision fork in figure ?.5>, the
choice is between HdrillK with a preference of >.=; and HdonBt drillK with a
preference of >.@>, so the choice is to drill. The preference of the act chosen is
written in pink at the base of each act fork in figure ?.5>. The act not chosen is
scored off and this is shown by the double bar in figure ?.5>.
<. #ontinue backwards through the tree, repeating steps 9 and ; until the base of the
tree is reached. "or instance, the preference of the decision to take the test is >.7<
&>.@>V>.=;U>.<V>.@', while the preference not to take the test is >.@=.
The analysis using preference theory therefore indicates the decision2makerBs best
strategy is to take the test and, if it gives a favourable result, drillF if it produces an
unfavourable result, do not drill.
!ith A+0, the decision2maker would be advised to drill immediately. The preference
theory approach takes into account the e*ecutiveBs natural conservatism and tells him
to take the seismic test first and drill only if it is favourable. The seismic test then is a
form of Hinsurance policyK, which is good for the conservative decision2maker in this
case, but not worth its price to the averages2player &who would always choose the
decision alternative that would ma*imise their A+0'.
65
"igure ?.7: #nalysis usin! &2C
R;<>,>>>
R<>,>>>
R9?>,>>>
Acquire seismic
data
R2;>,>>>
Test favourable
>.@
Test unfavourable
>.<
.rill
.onBt drill
%il
>.5
R<>>, >>>
/o oil >.6
.onBt acquire
seismic data
.rill
.onBt drill
%il
>.=?
R<>>,>>>
/o oil
>.5?
.rill
.onBt drill
%il
>.??
R<>>,>>>
/o oil
>.<?
R25>>,>>>
R<>>,>>>
R>
R5>>,>>>
R<>>,>>>
R>
R5>>,>>>
R<;>,>>>
R;>,>>>
R5;>,>>>
.)C
Act fork
Avent fork
:est strategy
)reen numbers A+0 of
decision
alternative
R25>>,>>>
R25>>,>>>
69
5
>.@
're"erence
R> R5>>>>> R9>>>>> R;>>>>> R<>>>>>
-et +i@uid Assets
"igure ?.6: )he decision-maker7s preference cur,e
Acquire seismic data
.onBt acquire
seismic data
R9<<,>>> &>.@V;<>>>>U5>>>>>V>.<'
R9?>,>>>
"igure ?.=: )est results eliminated
>.?
>
6;
"igure ?.5>: #nalysis usin! preferences
R25>>,>>>
R25>>,>>>
.)C
Act fork
Avent fork
$ed numbers (reference from curve
:est strategy
)reen numbers (reference for event fork
(ink numbers (reference for act
chosen
Act not chosen
>.7<
Acquire seismic
data
R2;>,>>>
>.7<
Test favourable
>.@
Test unfavourable
>.<
>.=;
>.@>
.rill
.onBt drill
>.5
%il
>.5
R<>>, >>>
/o oil >.6
.onBt acquire
seismic data
.rill
.onBt drill
>.=;
%il
>.=?
R<>>,>>>
/o oil
>.5?
>.@=
.rill
.onBt drill
>.@7
%il
>.??
R<>>,>>>
/o oil
>.<?
R25>>,>>>
R<>>,>>>,>.6=
R>,>.>
R5>>,>>>,>.@
R<>>,>>>,>.6=
R>,>.>
R5>>,>>>,>.@
R<;>,>>>,5.>>
R;>,>>>,>.97
R5;>,>>>,>.@=
6<
(reference theory can be e*tended to decisions involving multiple attributes. +ulti2
attribute preference theory &more commonly referred to as multi2attribute utility
theory' shows how, provided certain conditions apply, the main decision problem can
be broken into sub2problems and a single attribute preference function &or curve' can
be derived for each attribute and then these can be combined to obtain a multi2
attribute function &)oodwin and !right, 5665 p=@'. A number of methods have been
proposed for performing this analysis, but the approach described by 4eeney and
$aiffa &567@' is the most popular.
A number of researchers have questioned the application of preference theory to real
problems. +ost of their concerns relate to the generation of a decision2makerBs
preference curve. Since, crucially, whilst von /eumannn and +orgenstern proved
that a preference curve e*ists for each decision2maker who makes decisions consistent
with the eight a*ioms they did not specify how to obtain this curve.
Since 56<<, many researchers have studied this problem, but so far, their attempts
have been only marginally successful:
HThe unfortunate truth is that we &as a business community' do not as yet have
a satisfactory way to construct an individualBs preference curve.K
&/ewendorp, 566@ p5@9'
)enerally researchers have attempted to describe a decision2makerBs preference curve
by obtaining the decision2makerBs responses to a carefully designed set of
hypothetical investment questions &/ewendorp, 566@ p5@>F Dammond, 56@7F Swalm,
56@@'. In these tests, the decision2maker is offered a choice between a gamble having
a very desirable outcome and an undesirable outcome, and a no2risk alternative of
intermediate desirability. Such tests, though, have only been marginally successful
for two reasons. "irstly, the test procedures have to use hypothetical gambles and
decisions rather than actual gambles. The rationale for this is that if the procedure
used real decision2making situations, decision2makers generally would either accept
or re8ect a decision without stopping to e*plicitly state what the probabilities would
have to be for them to have been indifferent between the gamble and the no2risk
alternative &/ewendorp, 566@'. Tocher &5677' argues that since these gambles are
only imaginary, the decision2makerBs 8udgements about the relative attractiveness of
6?
the gambles may not reflect what the decision2maker would really do. This is known
as the preference reversal phenomenon and has been researched e*tensively &Slovic,
566?F +owen and )entry, 56=>, )rether and (lott, 5676F -ichenstein and Slovic,
5675F -indman, 5675' and many theories have been proposed to e*plain it &for
e*ample, %rdone1 and :enson, 5667F )oldstein and Ainhorn, 56=7'. Secondly, most
decision2makers are not used to making decisions on the basis of a precise
discernment of probabilities used to 8ustify a gamble. $ather, the probabilities are
usually specified for a given investment, and the parameter that the decision2maker
focuses on is whether the successful gain is sufficient to 8ustify the gamble
&/ewendorp, 566@'.
A further limitation of the application of preference theory to real problems is the
inability to construct corporate preference curves. Dammond &56@7' argues that an
organisationBs propensity for risk is higher than that of an individual, and that
therefore companiesB preference curves are usually more risk seeking than those of
individualsB. Dowever, Dammond believes, that the individual manager could
unwittingly apply his own much more conservative preference curve when making
decisions on behalf of the company. Therefore, Dammond &56@7' concludes that
organisations should have a corporate preference curve for individual mangers to use
when making company decisions. Dowever, whilst research into constructing
corporate preference curves has been attempted, its theoretical results have, to date,
been too comple* for practical application &Dammond, 56@7'. $ecently, though one
company based in Aberdeen, .et /orske 0eritas &./0', has started offering
preference theory2based analysis to upstream companies.
.espite these limitations, proponents of preference theory claim that it provides the
decision2maker with the most representative decision parameter ever developed.
They argue that its use will produce a more consistent decision policy than that which
results from using A+0 and that preference theory also accounts for the non2arbitrary
factors in an arbitrary way &Swalm, 56@@'. %thers are more cautionary. :ailey et al.
&in press' argue that proponents of preference theory should not claim that it is a
descriptive tool but rather offer it as a prescriptive technique that can be used to help
individuals or companies take decisions. They write:
6@
Hpreference theory does have a more limited but still important role. It can
graphically demonstrate to decision2makers what their style of decision2
making implies. It might show a highly conservative, very risk averse
decision2maker that there was room for much more fle*ibility without
incurring enormous penalties, or it might show an intuitive decision2maker
that his decision2making was altogether too risky.K
Dowever, opponents argue that it is impossible to quantify emotions regarding money,
and therefore the whole idea of preference theory is an e*ercise in frustration
&Tocher, 5677'.
3nderpinning preference theory, the A+0 concept, decision tree analysis and indeed
the whole of decision analysis, as highlighted in the sections above and in #hapter 9,
is the ability of the analyst to generate sub8ective probabilistic estimates of the
variables under investigation, as a mechanism for quantifying the risk and uncertainty.
Traditional probability theory relies on the relative frequency concept in which
probability is perceived to be the long2run relative frequency with which a system is
observed in a particular state in a series of identical e*periments. Dowever, given its
emphasis on repeated e*periments, the frequentist concept is unsuitable for modelling
under the conditions of risk and uncertainty present in the ma8ority of investment
decision problems. As such, the sub8ective probabilities used in decision analysis are
founded upon a quite a different conceptual base. As indicated in #hapter 9, to a
sub8ectivist, probability represents an observerBs degree of belief that a system will
adopt a particular state. There is no presumption of an underlying series of
e*periments. The observer need only be going to observe the system on one occasion.
+oreover, sub8ective probabilities encode something about the observer of the
system, not the system itself. The 8ustification for using sub8ective probabilities in
decision analysis does not 8ust rest on the case that frequentist probabilities are
inappropriate but also in the principles of consistency that the :ayesians suggest
should be embodied in rational decision2making &for a full discussion see "rench,
56=6 p;5'. Accepting the rationale that sub8ective probability estimates should be
used for investment decision2making under conditions of uncertainty however does
cause problems since, partly because of their sub8ective nature, there is no formula in
the decision analysis literature for generating these probabilities. Therefore, analysts
typically use their 8udgement, e*trapolate from historical data or, for e*ample when
estimating recoverable reserves for a particular field, use the results achieved in other
67
similar plays to guide their predictions. Traditionally analysts used single2value
probability estimates to e*press the degree of risk and uncertainty relating to the
uncertain parameters. +ore popular now is to generate sub8ective probability
estimates using risk analysis which adds the dimension of simulation to decision
analysis.
The following section draws on the prescriptive decision analysis literature first to
provide a brief overview of the main concepts of risk analysis and then to indicate the
impact of risk analysis on investment decision2making in the upstream. It is
important to recognise that risk analysis is a special case of decision analysis that uses
techniques of simulation. %ften in the literature, the terms are used interchangeably
leading to confusion when comparing accounts.
8(7 RI0. A-A+C0I0
Simulation as a means of risk analysis in decision2making was first applied to
petroleum e*ploration investments in 56@> &)rayson, 56@>'. The technique can be
applied to any type of calculation involving random variables. It can be used to
answer technical questions such as &H!hat is the volume of recoverable reserves of
hydrocarbons in this acreageIK' and economic ones such as &H!hat is the probability
that the /(0 of this prospect will e*ceed the target of R* millionIK' &:ailey et al., in
press'. The main concepts of risk analysis using simulation will now be presented
before its applicability to the upstream is e*amined.
$isk analysis based on +onte #arlo simulation is a technique whereby the risk and
uncertainty encompassing the main pro8ected variables in a decision problem are
described using probability distributions. Then by randomly sampling within the
distributions, many, perhaps thousands, of times, it is possible to build up successive
scenarios, which allow the analyst to assess the effect of risk and uncertainty on the
pro8ected results. The output of a risk analysis is not a single value, but a probability
distribution of all e*pected returns. The prospective investor is then provided with a
complete risk2return profile of the pro8ect showing all the possible outcomes that
could result from the decision to stake money on this investment.
6=
It is perhaps easiest to see how +onte #arlo simulation works by using an e*ample of
a hypothetical field. The main data is given in table ?.9. The decision facing the
decision2makers is whether to develop the field. (erforming a simple deterministic
calculation, with a discount rate of 5>E, gives an /(0 of R59? million and the
decision to go ahead on development should be straightforward.
:ut a probabilistic assessment of the same field gives the decision2maker a broader
picture to consider. Assume the probabilistic assessment uses the figures in table ?.9
as the Hmost likelyK inputs &those falling at the mid2point of the range' but also
suggests the ranges of possible values for inputs in table ?.;.
$eserves of 5?> million barrels of oil &+:>'
(roduction has a plateau &assumed to be reached immediately' of 59E per annum
of total reserves &i.e. 59E of 5?> +:%T5=+:%Nyr' for ? years, then declining at
9>E per year thereafter, until all 5?>+:% have been produced.
? production wells are needed, at a cost of R5?m per well over two years
(latformNpipeline costs are R7@?m over three years
Abandonment e*penditure is R;7? million after last production
%perating e*penditure is R7?million per year
#orporation ta* is ;>E
Inflation is ;.?E throughout the period
.iscount rate is 5>E
%il price assumed to be R5= per barrel rising at the rate of inflation
Table ?.9: Dypothetical field data
.rilling, capital and operating e*penditures are assumed to be HnormalK
distributions with a standard deviation &S.' of 5>E of the mean &S. is a measure of
the range of uncertainty'
Abandonment e*penditure is HnormalK, with S.T9>E of the mean
(roduction volumes are HnormalK, but with a positive correlation to operating
e*penditure
%il price is HlognormalK, with S.T5>E of the mean, in the first year of production
&9>><', rising by 9E per year, reaching ;<E by the last year of production. This
gives a roughly constant low oil price at about R5>Nbarrel, with the high price rising
from R9; to R;7.?Nbarrel through field life.
Table ?.;: Dypothetical field data for 2onte Carlo simulation
Ten thousand +onte #arlo trials give the results shown in table ?.<. The mean, or
average, or e*pected value is R59< million &that is, a statistically significant number of
identical opportunities would, on average, be worth R59< million, in /(0 terms.'.
66
'ercentile ?alue
>
5>
9?
?>
7?
6>
5>>
2559
97
75
599
57@
99;
<99
Table ?.< : <esults from the 2onte Carlo simulation
:ut in fact there is a range of possible outcomes and a chance of very different results.
"or e*ample, the so2called p5> value, or forecast with 5>E possibility of occurrence,
&see table ?.<' is R97 million, so 5>E of the cases run in the simulation gave values
less than R97 million. The lowest possible outcome is R2559+ and ?E of the cases, or
trials, gave negative /(0s. %n the other hand, the p6> was R99; million, so 5>E of
the trials gave values greater than R99; million &:ailey et al., in press'.
"or this particular field, there is a small, but not 1ero &that is, appro*imately <E'
chance of losing money. The decision would probably still be to go ahead, but the
+onte #arlo analysis, by revealing the wider picture, gives the decision2makers
greater comfort that their decision has taken everything into account.
3sing risk analysis in investment appraisal has a number of advantages. "irstly and
most importantly, it allows the analyst to describe risk and uncertainty as a range and
distribution of possible values for each unknown factor, rather than a single, discrete
average or most likely value. #onsequently when +onte #arlo simulation is used to
generate a probability distribution of /(0, /ewendorp &566@ p;7?' believes that:
HThe resulting profit distribution will reflect all the possible values of the
variable.K
This is a slightly dubious claim since the resulting profit distribution will not contain
every possible value of /(0. It will only include those that the decision2maker or
analyst feels are likely to occur. There is always the possibility of Hacts of )odK or
Htrain wrecksK &see section ?.7 and for a full discussion refer to Spencer and +organ,
5>>
566='. Dowever, it is certainly true, that in generating probabilistic output, the
decision2maker is more likely to capture the actual value in the predicted range.
Secondly, risk analysis allows the analyst to identify those factors that have the most
significant effect on the resulting values of profit. The analyst can then use
sensitivity analysis to understand the impact of these factors further. There are
several ways this sensitivity analysis can be carried out and the reader is
referred to Singh and 4inagi &56=7' for a full discussion.
Implementing risk analysis using +onte #arlo simulations has limitations and
presents a number of challenges. "irstly, +onte #arlo simulations do not allow for
any managerial fle*ibility. This can be overcome by running simulations for several
scenarios &)alli et al., 5666'. )utleber et al. &566?' present a case study where
simulations were carried out to compare three deals involving an oil company and
local government. +urtha &5667' provides many references to practical applications
of this procedure. Secondly, whilst geologists intuitively e*pect to find a correlation
between, for e*ample, hydrocarbon saturation and porosity, this is not acknowledged
e*plicitly in the literature nor are analysts given any guidance concerning how to
model such relationships.
)oodwin and !right &5665 pp5?;25?7' describe types of dependence and approaches
to modelling dependence are given in /ewendorp &566@ pp<>@2<;5', )oodwin and
!right &5665 pp5?;25?7', Ailon and "awkes &567;' and Dull &5677'. There is
significant evidence in the prescriptive decision analysis literature that decision2
makers have difficulty assigning the strength of association between variables.
/isbett and $oss &56=> p9@' have given the following concise summary of this
literature:
HThe evidence shows that people are poor at detecting many sources of
covariation (erception of covariation in the social domain is largely a
function of pre2e*isting theories and only very secondarily a function of true
covariation. In the absence of theories, peopleBs covariation detection
capacities are e*tremely limited. Though the conditioning literature shows
that both animals and humans are e*tremely accurate covariation detectors
under some circumstances, these circumstances are very limited and
constrained. The e*isting literature provides no reason to believe that
humans would be able to detect relatively weak covariations among stimuli
5>5
that are relatively indistinctive, subtle and irrelevant motivationally and, most
importantly, among stimuli when the presentation interval is very large.K
#hapman and #hapmanBs 56@6 study provided evidence of a phenomenon that they
refer to as illusory correlation. In their e*periment, naXve 8udges were given
information on several hypothetical mental patients. This information consisted of a
diagnosis and drawing made by the patient of a person. -ater the 8udges were asked
to estimate how frequently certain characteristics referred to in the diagnosis, such as
suspiciousness, had been accompanied by features of the drawing, such as peculiar
eyes. It was found that 8udges significantly overestimated the frequency with which,
for e*ample, suspiciousness and peculiar eyes occurred together. +oreover, this
illusory correlation survived even when contradictory evidence was presented to the
8udges. Tversky and 4ahneman &567<' have suggested that such biases are a
consequence of the availability heuristic. It is easy to imagine a suspicious person
drawing an individual with peculiar eyes, and because of this, the real frequency with
which the factors co2occurred was grossly overestimated. So, in the case of the
relationship between porosity and water saturation, this research suggests that because
geologists e*pect there to be a correlation, if there is any evidence of a correlation in
any particular case, the geologist is likely to overestimate the strength of this
relationship. This research indicates the powerful and persistent influence that
preconceived notions can have on 8udgements about relationships &)oodwin and
!right, 5665 p5?;'.
The third limitation of +onte #arlo simulations is perhaps most significant. In the
industry literature, no published study has indicated which probability distribution
most accurately describes the reservoir parameters of reservoir rocks of similar
lithology and water depth. Similarly, there has been no research that has identified the
appropriate shape of probability distribution to be adopted for economic factors such
as oil price. Section @.9 of #hapter @ will discuss how companies cope with this lack
of prescription in the literature. It is possible here, to perform a crude test to
investigate whether the shape of the probability distribution used for each input
variable, affects the estimate generated by a +onte #arlo simulation. Such a test is
carried out below.
5>9
"or each reservoir parameter, base values are entered and probability distributions are
assigned to each of these variables from the seventeen available in #rystal :allY .
Then a +onte #arlo simulation is run and the estimate of recoverable reserves
generated e*pressed in percentiles, is noted. This process is repeated twelve times
altering only the probability distribution assigned to each variable each time. The
base value data and the probability distributions used for each trial are shown in table
?.?. The output produced is summarised in table ?.@ and provides evidence that
altering the probability distribution assigned to each reservoir parameter, significantly
affects the forecast of the recoverable reserves. &/ote, that although some of the
distributions used here are more unusual &for e*ample, the !eibull', the lack of
prescription in the literature over the shape of probability distribution that analysts
should adopt for reservoir parameters &and economic variables' means that, if these
results are accurate, analysts could, unwittingly or otherwise, use these types of
distribution to distort the results.'
"urther studies are needed to confirm these results. This would then prompt
researchers to e*plore the shape of the probability distributions to be used for the
reservoir parameters of reservoir rocks of similar lithology and burial history and the
nature of the probability distributions to be used to model the economic variables.
"uture research should also investigate the nature of correlation between the reservoir
&and economic' variables. The author of this thesis tried repeatedly throughout the
course of the study to access HrealK reserves and economic data to conduct such
research but was unable to collect enough data to make any results achieved
meaningful. !hilst much of the economic data is regarded by companies as
commercially sensitive and this makes such research unlikely in the near future, a
book due for publication ne*t year should contain the relevant reserves data &)luyas
et al., 9>>5'. It is hoped that the interest provoked by this thesis will motivate
researchers and practitioners to conduct the necessary studies.
.espite these limitations, risk analysis using simulation is perceived by the ma8ority
of decision analysts to enable a more informed choice to be made between investment
options &for e*ample, /ewendorp, 566@F )oodwin and !right, 5665 p5?5'.
#ertainly, by restricting analysts to single2value estimates the conventional /(0
approach yields no information on the level of uncertainty that is associated with
5>;
different options. Despos and Straussman &56@?' have shown how the simulation
approach can be e*tended to handle investment problems involving sequences of
decisions using a method known as stochastic decision tree analysis.
,ype o" distri#ution assigned to reach reservoir parameter "or each run
Reservoir
parameters
1ase
value
Rec Res 3 Rec Res % Rec Res 5 Rec Res 7 Rec Res 8 Rec Res :
)$0 >.9 /ormal /ormal /ormal /ormal /ormal /ormal
/N) >.= Triangular Triangular Triangular 3niform -ognormal -ognormal
(orosity >.;9 Triangular -ognormal Triangular 3niform -ognormal -ognormal
Shc >.= Triangular Triangular Triangular Triangular -ognormal -ognormal
$e >.<? Triangular Triangular -ognormal Triangular -ognormal -ognormal
Rec Res 9 Rec Res ; Rec Res 6 Rec Res 3& Rec Res 33 Rec Res 3%
)$0 >.9 /ormal -ognormal -ognormal !eibull /ormal :eta
/N) >.= Triangular !eibull 3niform !eibull :eta !eibull
(orosity >.;9 !eibull :eta 3niform !eibull (areto Triangular
Shc >.= Triangular 3niform 3niform !eibull 3niform -ognormal
$e >.<? Triangular Triangular 3niform !eibull !eibull !eibull
Table ?.?: .ase ,alue data and pro'a'ility distri'utions assi!ned to each of the reser,oir parameters
(-<CE!ross rock ,olume, @?-Enet to !ross, ShcEhydrocar'on saturation, <eEreco,ery efficiency,
<ec <esEreco,era'le reser,es
'ercentile Rec
Reserves
3
Rec
Reserves
%
Rec
Reserves
5
Rec
Reserves
7
Rec
Reserves
8
Rec
Reserves
:
>E 5>7 6; 67 6? 6< 76
3&K 35; 35% 35% 35% 359 3%&
9>E 5<@ 5<; 5<9 5<9 5<? 5;;
;>E 5?9 5?> 5?> 5?5 5?; 5<5
<>E 5?7 5?7 5?@ 5?= 5?= 5?>
8&K 3:5 3:% 3:5 3:7 3:7 3:&
@>E 5@7 5@6 5@6 579 57> 57>
7>E 57; 57@ 577 5=> 57@ 5=5
=>E 5=> 5=? 5== 5== 5=< 56@
6&K 3;6 36; %&5 %&5 36: %39
5>>E 99= 9?? 9@; 9=5 9@> ;;6
'ercentile Rec
Reserves
9
Rec
Reserves
;
Rec
Reserves
6
Rec
Reserves
3&
Rec
Reserves
33
Rec
Reserves
3%
>E 5?> ? 5>? 9;?6 57 6
3&K 537 77 35% 33866 %7: 335
9>E ;66 @@ 5<5 5@@?5 ;=9 57?
;>E <@7 =< 5<6 956?? ?97 99?
<>E ?9; 5>; 5?7 979>6 @7? 9=>
8&K 86; 3%8 3:5 5%;7% ;39 559
@>E @?7 5<= 57> <>=@@ 6=< <><
7>E 79@ 57@ 57= <=@?> 59>@ <6=
=>E =57 9>? 5=7 @96@@ 5?<< @9@
6&K 693 %8& %&3 ;;;88 %%95 ;36
5>>E 56;= ?<@ 9=7 9;76>6 575<; 96>;
Table ?.@: )a'le of the output !enerated usin! the 'ase ,alue data and input distri'utions specified in
)a'le 4.4 (<ec reser,es E reco,era'le reser,es
5><
The ne*t section draws on the decision theory and industry literatures to present
portfolio theory, a technique that has been used within the finance industry for a
number of years but which has only recently been applied to petroleum investment
decisions. Therefore, the concepts of portfolio theory will be outlined first before its
applicability to upstream investment decision2making is analysed.
8(8 'OR,FO+IO ,)ORC
In practice, a business will normally invest in a range, or portfolio, of investment
pro8ects rather than in a single pro8ect. The problem with investing all available funds
in a single pro8ect is, of course, that an unfavourable outcome could have disastrous
consequences for the business. :y investing in a spread of pro8ects, an adverse
outcome from a single pro8ect is unlikely to have ma8or repercussions. Investing in a
range of different pro8ects is referred to as diversification, and by holding a diversified
portfolio of investment pro8ects, the total risk associated with the business can be
reduced &Atrill, 9>>> p5=?'. This introduces two concepts. "irst, asset value is
additive. The incremental e*pected value that an asset adds to the portfolioBs
e*pected value is the assetBs e*pected value. Second, asset risk is not additive. The
amount of risk an asset contributes to the portfolio is not solely dependent on its risk
as a stand2alone investment &measured in finance theory by the standard deviation of
the e*pected value probability distribution' &!hiteside, 5667'. Atrill &9>>> p55='
e*plains this by dividing the total risk relating to a particular pro8ect into two
elements: diversifiable risk and non2diversifiable risk &figure ?.55':
.iversifiable risk is that part of the total risk which is specific to the pro8ect, such
as reserves, changes in key personnel, legal regulations, the degree of competition
and so on. :y spreading the available funds between investment pro8ects, it is
possible to offset adverse outcomes in occurring in one pro8ect against beneficial
outcomes in another &Atrill, 9>>> p5=='.
/on2diversifiable risk is that part of the total risk that is common to all pro8ects
and which, therefore, cannot be diversified away. This element of risk arises from
general market conditions and will be affected by such factors as rate of inflation,
5>?
the general level of interest rates, e*change rate movements and so on &Atrill,
9>>> p5=='. Arguably, the most critical non2diversifiable risk for e*ploration
companies is the oil price.
It then follows that the incremental risk an asset adds to the portfolio will always be
less than its stand2alone risk.
Total risk
Risk
.iversifiable
risk

-um#er o" assets
"igure ?.55: <educin! risk throu!h di,ersification (source: Di!son, 1994 p159
There are two types of diversification: simple and +arkowit1. Simple diversification
&commonly referred to as market or systematic diversification in the stock market'
occurs by holding many assets. It holds that if a company invests in many
independent assets of similar si1e, the risk will tend asymptotically towards 1ero. "or
e*ample, as companies drill more e*ploration wells, the risk of not finding oil reduces
towards 1ero. #onsequently, companies that endorse a strategy of taking a small
equity in many wells are adopting a lower risk strategy than those that take a large
equity in a small number of wells. Dowever, the economic returns on independent
assets are to, a greater or lesser e*tent, dependent on the general economic conditions
and are non2diversifiable. 3nder these conditions, simple diversification will not
reduce the risk to 1ero but to the non2diversifiable level. +arkowit1 diversification
relies on combining assets that are less than perfectly correlated to each other in order
to reduce portfolio risk. The method is named after a 566> /obel (ri1e recipient, a
financial theorist, who first introduced the technique in his 56?9 paper entitled
%ortfolio Selection. +arkowit1 diversification is less intuitive than simple
diversification and uses analytical portfolio techniques to ma*imise portfolio returns
/on2diversifiable risk
5>@
for a particular level of risk. This approach also incorporates the fact that assets with
low correlation to each other when combined have a much lower risk relative to their
return &!hiteside, 5667'.
3sing these principles, portfolio optimisation is a methodology from finance theory
for determining the investment program and asset weightings that give the ma*imum
e*pected value for a given level of risk or the minimum level of risk for a given
e*pected value. This is achieved by varying the level of investment in the available
set of assets. The efficient frontier is a line that plots the portfolio, or asset mi*,
which gives the ma*imum return for a given level of risk for the available set of
assets. (ortfolios that do not lie in the efficient frontier are inefficient in that for the
level of risk they e*hibit there is a feasible combination of assets that result in a
higher e*pected value and another which gives the same return at lower risk. &/ote,
in reality, due to real world constraints such as the indivisibility of assets, trading
costs and the dynamic nature of the world, all practical portfolios are inefficient'.
To calculate the efficient frontier it is imperative to determine the mean return of each
asset &usually the A+0 in industrial applications', the variance of this value &defined
as risk in finance theory' and each assetBs correlation to the other assets in the
available set of investments &!hiteside, 5667'. This classification of risk assumes
that:
"irmsB long run returns are normally distributed and can, consequently, be
adequately defined in terms of the mean and variance. In reality, it is likely that
the distribution describing long run returns would be HskewedK.
0ariance is a useful measure of risk. In calculating variance, positive and negative
deviations from the mean are equally weighted. In fact, decision2makers are often
more pre2occupied with downside risk J the risk of failure. A solution to this
problem is to determine the efficient set of portfolios by using another risk
measure. A group of suitable risk measures that only takes the dispersions below
a certain target into account are the downside risk measures. In the mean
downside risk investment models the variance is replaced by a downside risk
measure then only outcomes below a certain point contribute to risk.
5>7
There is enough information to estimate the mean and variance of the distribution
of outcomes. This does require a high level of information that, in some cases, is
not available &$oss, 5667'.
Dowever, provided that the assumption that variance is a useful appro*imation of risk
is accepted, the aim is to ma*imise the e*pected return under a certain level of
variance, which is equivalent to minimising the variance under a certain level of
e*pected return. To determine the variance, +onte #arlo simulation is used. "irst,
information about all the variables that affect the calculation of a cash flow of one of
the pro8ects is collected and their probability distributions are estimated. Then for
each pro8ect, using +onte #arlo simulation, a number of cumulative discounted cash
flows and their matching discounted investments can be generated simultaneously.
"rom these points, the A+0 of each pro8ect, the variance of the A+0 of each pro8ect
and the correlation between the A+0s of the different pro8ects can be calculated. :y
simulating the cash flows of all pro8ects simultaneously, some of the uncertain
variables, which fi* the systematic risk and thus provide for the correlation between
the different pro8ects, are equal for all pro8ects and therefore the coefficient of
correlation between pro8ects can be determined. &The value of this coefficient can
range from U5 in the case of perfect correlation, where the two assets move together,
to J5 in the case of perfect negative correlation, where the two assets always move in
opposite directions. The coefficient is > when there is no association between the
assets and they are said to be independent.' Then these values, together with any
other constraints, are used to generate the efficient frontier. To find the efficient
portfolios, +arkowit1 defined the mean variance model that reduces to a quadratic2
programming problem that is easily solved by the many mathematical software
packages available. After the efficient set has been determined, a portfolio can be
chosen from this group. There are several ways of doing this. "or the method
advocated by +arkowit1 &and stochastic dominance' the utility function, or
preference curve, of the company would need to be determined, as discussed above in
section ?.;, this is particularly difficult. Therefore some finance theorists advocate
the use of one of the safety2first criteria &for a full discussion see $oss, 5667'.
5>=
Investment opportunities of the oil industry have a great resemblance to financial
assets. As with the financial assets, there is much risk and uncertainty about the profit
of the pro8ects. Assets are highly correlated with each other. They all have some
variables, such as oil price, that affect the profitability of the pro8ect in common. In
the stock market paper assets &stocks and shares' are traded and in the oil business
companies hold and trade portfolios of real assets by, for instance buying and selling
shares in 8oint ventures.
The following simple e*ample from :all and Savage &5666' shows how the
application of the principles of portfolio theory to the oil industry can result in
decisions that are counter2intuitive.
An oil company has R5> million to invest in e*ploration and production pro8ects.
%nly two pro8ects are available and each requires the full R5> million for 5>>E
interest. %ne pro8ect is relatively HsafeKF the other relatively HriskyK. The chances of
success are independent. The facts about the pro8ects are presented in table ?.7.
Outcome -'?
!Fmillion$
Independent
'ro#a#ility !K$
0AF) .ry hole 25> <>
Success ?> @>
RI0.C .ry hole 25> @>
Success => <>
Table ?.7: Safe and risky pro$ects (source: .all and Sa,a!e, 1999
The A+0s of each pro8ect are the same:
A+0
safe
T@>EVR?>U<>EV&R25>'TR9@ million
A+0
risky
T<>EVR=>U@>EV&R25>'TR9@ million
A complication is now added. If money is lost, shareholder confidence is forfeited.
There is a <>E chance of forfeiting shareholder confidence with the safe pro8ect, and
a @>E chance with the risky pro8ect. Since the A+0 for both pro8ects is R9@ million,
5>6
there is no way of increasing that by choosing the risky over the safe pro8ect. 3nder
both circumstances the safe pro8ect is obviously the better choice.
A further complication is added. Suppose it is possible to split the investment evenly
between the two pro8ects. Intuitively it would seem a bad idea to take a ?>E out of
the safe pro8ect and put it into the risky one. Dowever, intuition is not always the best
guide.
There are now four possible outcomes and these are shown in table ?.=. The A+0 of
portfolio is still R9@ million &9<E VR@?U;@EVR9>U5@EVR;?U9<EV&2R5>'TR9@
million' but the only way to forfeit shareholder confidence is to drill two dry wells
&Scenario <', for which the probability is 9<E. That cuts the risk of forfeiting
shareholder confidence by almost half. So, moving money from a safe pro8ect to a
risky one, which, of course, seems counter2intuitive, reduces risk and is the effect of
diversification.
0*)-ARIO 0AF) RI0.C 'RO1A1I+I,C R),UR-!Fmillion$ R)0U+,
3
%
5
7
Success
Success
.ry
hole
.ry
hole
Success
.ry
hole
Success
Success
>.@V>.<T>.9<
>.@V>.@T>.;@
>.<V>.<T>.5@
>.<V>.<T>.9<
?>EVR?>U?>EVR=>TR@?
?>EVR?>U?>EV&R25>'TR9>
?>EV&R5>'U?>EV&R=>'TR;?
?>EV&2R5>'U?>EV&2R5>'T2R5>
Shareholder
confidence retained
Shareholder
confidence retained
Shareholder
confidence retained
Shareholder
confidence lost
Table ?.=: #ll possi'le outcomes of in,estin! 49F in each pro$ect (source: .all and Sa,a!e, 1999
+ost companies that do not use portfolio theory rank their e*ploration pro8ects based
on A+0 and then choose the pro8ect with the highest A+0 &Section @.9 of #hapter
@'. This ignores the diversification effect and in the e*ample above would have led to
allocating all the funds to the safe pro8ect, with nearly twice the risk of the best
portfolio &:ailey et al., in press'.
(ublications from !hiteside &5667' and $oss &5667' provide further details of how
portfolio theory can be applied to the industry. Software companies such as +erak
and Indeva produce tools that allow upstream companies to use the technique easily.
55>
$ecently the application of another technique from finance theory, option theory, has
been gaining attention in the literature as a tool for valuing undeveloped oil reserves.
Dowever, currently the discussion raises more questions than it answers, and the
method has yet to be shown to be a viable method for evaluating these reserves &see,
for e*ample, -ohren1 and .ickens, 566;F +arkland, 5669'. The following section
reviews the industry and decision theory literature on option theory.
8(: O',IO- ,)ORC
The application of option theory to investment appraisal was motivated by a
recognition that the standard .#" approach does not capture all sources of value
associated with a given pro8ect. Specifically, writers such as .i*it and (indyck
&566<' argue that two aspects of e*tra value or economic desirability are inadequately
captured by a standard /(0 analysis. "irst, the operating fle*ibility available within
a single pro8ect, which enables management to make or revise decisions at a future
time. The traditional /(0 method, they believe, is static in the sense that operating
decisions are viewed as being fi*ed in advance. In reality, :uckley &9>>> p<99'
argues, good managers are frequently good because they pursue policies that maintain
fle*ibility on as many fronts as possible and they maintain options that promise
upside potential. "ollowing such an observation .i*it and (indyck &566< p@' write:
Hthe ability to delay an irreversible investment e*penditure can profoundly
affect the decision to invest. It also undermines the simple net present value
rule, and hence the theoretical foundation of standard neoclassical investment
models.K
They go on to conclude:
Has a result the /(0 rule must be modified.K &.i*it and (indyck, 566<
p@'
Secondly, they believe that the HstrategicK option value of a pro8ect, which results
from its interdependence with future and follow2up investments, is not accounted for
in the conventional /(0 method &.i*it and (indyck, 566= and 566<'. Therefore,
+yers &56=<' and 4ester &56=<' suggest that the practice of capital budgeting should
555
be e*tended by the use of option valuation techniques to deal with real investment
opportunities.
%ption theory, sometimes called Hoption pricingK, Hcontingent claims analysisK or
Hderivative asset evaluationK, comes from the world of finance &-ohren1 and .ickens,
566;'. In its most common form, option theory uses the :lack2Scholes model for
spot prices and e*presses the value of the pro8ect as a stochastic differential equation
&)alli et al., 5666'. In this section, by reviewing the finance literature, the
development of option theory will be traced. The popularity and success of option
theory algorithms has led to wide interest in analogous application to evaluation of oil
and gas assets &-ohren1 and .ickens, 566;'. This literature will also be reviewed.
In the 567>s, the financial world began developing contracts called puts and calls.
These give the owner the right, for a fee, to buy an option, which is the right &but not
the obligation' to buy or sell a financial security, such as a share, at a specified time in
the figure at a fi*ed price &:ailey et al., in press'. If the transaction has to take place
on that date or never, the options are called AuropeanF otherwise they are called
American &this does not refer to where the transaction takes placeZ' &)alli et al.,
5666'. An option to buy is known as a call option and is usually purchased in the
e*pectation that the price of the stock will rise. Thus a call option may allow its
holder to buy a share in company A:# for R?>> on or before ,une 9>>5. If the price
of the stock rises above R?>> the holder of the option can e*ercise it &pay R?>>' and
retain the difference. The holderBs payoff is that sum minus the price paid for the
option. A put option is bought in the e*pectation of a falling price and protects
against such a fall. The e*ercise price is the price at which the option can be
e*ercised &in this case R?>>' &:ailey et al., in press'.
The central problem with options is working out how much the owner of the contract
should pay at the outset. :asically, the price is equivalent to an insurance premiumF it
is the e*pected loss that the writer of the contract will sustain. #learly, the ability to
e*ercise the option at any time up to the maturity date makes American options more
valuable than Auropean options. !hat is less obvious is that this apparently minor
difference necessitates different procedures for calculating option prices &)alli et al.,
5666'.
559
The standard assumption made in option theory is that prices follow lognormal
:rownian motion. :lack and Scholes developed the model in the early 567>s.
A*perimental studies have shown that it is a good appro*imation of the behaviour of
prices over short periods of time. If they obey the standard :lack2Scholes model,
then the spot price, p
sp
, satisfies the partial differential equation:
d p
sp
T p!
t
U p
sp
dt
where p
sp
Tspot price, !
t
T:rownian motion, Tdrift in spot prices, pTprice and
Tspot price volatility.
Applying ItoBs -emma, the e*plicit formula for p
sp
can be shown to be:
p
sp
T p
>
e
[L& 2 9'N9M t U!t\
where p
sp
Tspot price, !
t
T:rownian motion, Tdrift in spot prices, and Tspot price
volatility, p
>
Toil price, tTtime
It is relatively easy to evaluate Auropean options &which can be e*ercised only on a
specified date', particularly when the prices follow :lack2Scholes model because
there is an analytic solution to the corresponding differential equation. This gives the
e*pected value of ma*Lp e*p &2rt' J p
k
, >M for a call or of ma*Lp
k
J p e*p&2rt',>M for a
put. In general the simplest way of getting the histogram and the e*pected value is by
simulating the diffusion process and comparing each of the terminal values to the
strike price of the option. Solving the partial differential equation numerically gives
only the option price &)alli et al., 5666'.
Avaluating the price of an American option is more difficult because the option can be
e*ercised at any time up to the maturity date. "rom the point of view of stochastic
differential equations, this corresponds to a free2boundary problem &see !ilmot,
.ewynne and Dowison, 566<'. The most common way of solving this type of
55;
problem is by constructing a binomial tree. This is similar to a decision tree &for an
e*planation see )alli et al., 5666'.
In the formula for /(0 given in section ?.9, the discount rate was used to account for
the effect of time on the value of moneyF however, this is not immediately apparent in
the option pricing equations. In option theory, the time value of money is
incorporated through the risk2free rate of return and by way of a Hchange of
probabilityK &Smith and +c#ardle, 5667, :a*ter and $ennie, 566@ and Trigeorgis,
566@, all provide good e*planations of this' &)alli et al., 5666'.
The application of these methods to HrealK, as opposed to financial options, dates back
to +yers &5677' and was popularised by +yers &56=<' and 4ester &56=<' &see +ason
and +erton &56=?' for an early review and .i*it and (indyck &566<' for a survey of
the current state of the art'. In this approach rather than determining pro8ect values
and optimal strategies using sub8ective probabilities and utilities, the analyst seeks
market2based valuations and policies that ma*imise these market values. In
particular, the analyst looks for a portfolio of securities and a trading strategy that
e*actly replicates the pro8ectBs cash flows in all future times and all future states. The
value of the pro8ect is then given by the current market price of this replicating
portfolio. The fundamental principal underlying this approach is the Hno arbitrageK
principle or the so2called Hlaw of one priceK: two investments with the same payoffs
at all time and in all states J the pro8ect and the replicating portfolio must have the
same value.
The idea of investments as options is well illustrated in the decision to acquire and
e*ploit natural resources. The similarity of natural resources to stock market options
is obvious. Stock market options give the holder the right but not the obligation to
acquire or sell securities at a particular price &the strike price' within a specified
timeframe but there is not an obligation to do so. The owner of an undeveloped oil
well has the possibility of acquiring the proceeds from the oil wellBs output but does
not have an obligation to do so and the company may defer selling the proceeds of the
assetBs output. "urther, much as a stock pays dividends to its owner, the holder of
developed reserves receives production revenues &net of depletion'. Table ?.6 lists the
important features of a call option on a stock &or, at least, all those necessary to enable
55<
one to price it' and the corresponding aspects of the managerial option implicit in
holding an undeveloped reserve &Siegel et al., 56=7'.
3sing this analogy, :rennan and Schwart1 &56=?' worked out a way to e*tend it to
valuing natural resource pro8ects using #hilean copper mines to illustrate the
procedure. They reasoned that managerial fle*ibility should improve the value of the
pro8ect. They allowed for three options: production &when prices are high enough',
temporary shutdown &when they are lower' and permanent closure &when prices drop
too low for too long'. .ifferent costs were associated with changing from one
production option to another. They found the threshold copper prices at which it was
optimal to close a producing mine temporarily &)alli et al., 5666'.
0,O*. *A++ O',IO- U-D)?)+O')D R)0)R?)0
#urrent Stock (rice #urrent value of developed reserves
A*ercise price .evelopment cost
Time to e*piration $elinquishment requirement
$iskless rate of interest $iskless rate of interest
.ividend /et production revenue less depletion
Table ?.6: )he similarities 'et+een a stock call option and unde,eloped reser,es (source: %addock et
al., 19AA
In practice, the key to applying options is in defining the options that are actually
available to management. Trigeorgis &566@' lists a whole range of managerial options
covering research and development and capital intensive industries, as well as oil and
mining. .i*it and (indyck &566<', the other classic te*t on real options, describes
several oil applications, including sequencing decision2making for opening up oil
fields and a study on building, mothballing and scrapping oil tankers. Since :rennan
and Schwart1Bs seminal work, many others have studied petroleum options.
#opeland, 4oller and +urrin &566>' for e*ample, describe a case involving an option
to e*pand production.
$eal option theory is best illustrated by an e*ample. The following illustration is
taken from -eslie and +ichaels &5667'.
Suppose an oil company is trying to value its license in a block. (aying the license
fee is equivalent to acquiring an option. The company now has the right &but not the
55?
obligation' to invest in the block &at the e*ercise price' once the uncertainty over the
value of the developed reserves &the stock price' has been resolved.
Assume that the company has the opportunity to acquire a five2year license and that
the block is e*pected to contain some ?> million barrels of oil. The current price of
oil from the field in which the block is located is R5> per barrel and the cost of
developing the field &in present value terms' is R@>> million. 3sing static /(0
calculations the /(0 will be R?>> million 2 R@>> millionTR25>> million.
The /(0 is negative so the company would be unlikely to proceed. The /(0
valuation ignores the fact that decisions can be made about the uncertainty, which in
this case is twofoldF in the real world there is uncertainty about the quantity of oil in
the block and about its price. It is, however, possible to make reasonable estimates of
the quantity of oil by analysing historical data in geologically similar areas and there
is also some historical data on the variability of oil prices.
Assume that these two sources of uncertainty between them result in a ;>E standard
deviation around the growth rate of the operating cash inflows. Assume also that
holding the option obliges the company to incur the annual fi*ed costs of keeping the
reserve active, say R5? million. This represents a dividend2like payout of ;E
&5?N?>>' of the value of the asset.
3sing the :lack and Scholes formula for valuing a real option
$%0TSe
2t
V [&/&d
5
'\ 2 We
2rt
V [&/&d
9
''
where, d
5
T[5n&SNW'U&r2U
9
N9't\N V t, d
9
T d
5
2 V t, STpresented value of e*pected
cash flows, WTpresent value of fi*ed costs, Tthe value lost over the duraction of the
option, rTrisk free interest rate, Tuncertainty of e*pected cash flow and tTtime to
e*piry,
and substituting for the values in this e*ample:
55@
$%0T&?>>e
2>.>;V?
' V [&>.?='\ J &@>>e
2>.>?V?
' V [&>.;9'\ T R9?5 million 2 R5?5 million T
R5>> million.
The R9>> million difference between the /(0 valuation of R25>> million and the
$%0 valuation of R5>> million represents the value of the fle*ibility brought about by
having the option to wait and invest when the uncertainties are resolved.
"rom a theoretical point of view, the key to applying real option theory is deciding
which variable is assumed to follow a :lack and Scholes model. :rennan and
Schwart1 assume that the spot price &here the oil price' obeys this model. Trigeorgis
&566@', 4emma &566;' and (addock et al. &56==' show a radically different approach
in which the analysis is based on the hypothesis that the pro8ect itself obeys this
model. The difference is important because the theory of option pricing requires a
liquid market for the underlying commodity, no transaction costs and no arbitrage.
!hile this is probably true for oil prices, it is doubtful whether a large enough market
e*ists for oil pro8ects &)alli et al., 5666'.
#oncerns have been e*pressed about all these approaches, usually directly
questioning the underpinning assumptions of the :lack and Scholes methodology and
its appropriateness for valuing real options particularly those with long time hori1ons
&for e*ample, -ohren1 and .ickens, 566;'. :uckley &9>>>' bypasses these criticisms
by describing an alternative route to valuing real options involving a decision tree
approach.
%ption theory methods are heralded as an improvement over traditional .#" methods
specifically because they allow managerial fle*ibility to be modelled and included in
the investment analysis. Dowever, since the value of an option is, in fact, an
e*pectation or, more precisely, the conditional e*pectation of the value given the
initial conditions, real options, like decision trees, do not give any indications about
the uncertainty of the pro8ect &)alli et al., 5666'. +ore importantly, a number of
professional managers have suggested that while the analogy relating managerial
fle*ibility to options has intuitive appeal, the actual application of option based
techniques to capital budgeting is too comple* &or certainly more comple* than the
/(0 method' for practical application &see #hapter @'.
557
This section and the ones that precede it &?.92?.?' have provided an overview of the
decision analysis techniques available to petroleum e*ploration companies to utilise in
their investment appraisal decision2making. All the tools described have been applied
to upstream investment analysis in the literature, allow risk and uncertainty to be
quantified and, crucially, are complementary. They do not represent alternatives.
This is important since, as indicated above, each tool has its limitations, so that
reliance only on the output of one tool for investment decision2making would be
inadvisable. :y combining the output from a variety of tools, the decision2maker is
more likely to assess the risk and uncertainty accurately. The tools described in the
sections above use similar input and, hence their use together does not place
unnecessary strain on the resources of an organisation. There are other techniques
described in the literature &for e*ample, the analytic hierarchy process &Saaty, 56=>'
and +arkov chain analysis &.as et al., 5666' but these have either not been applied to
the upstream or the input they demand and, in many cases, the output they produce, is
not complementary to the other investment techniques used by organisations. Dence
their use would represent a significant amount of additional work for the
organisations. "or these reasons, the tools described in the sections above, the
researcher believes comprise the toolkit currently available to the upstream decision2
maker.
The following section provides an indication of how these tools can be integrated into
one approach for investment appraisal in the upstream. There are numerous other
ways that the tools can be combined. The main aim of the ne*t section is to
demonstrate that the tools are complementary.
8(9 *URR)-, *A'A1I+I,C
The techniques presented above represent current theory in investment appraisal
decision2making in the upstream oil and gas industry. This section presents an
illustration of how these tools can be used together when an upstream company is
considering whether to drill an e*ploration well in a virgin basin at an estimated cost
of S5> million. It has been informed, modified and validated using knowledge gained
from the decision theory and oil industry literatures and insights ascertained from
55=
attendance at conferences and seminars during the course of the research. The
approach is summarised in figure ?.59.
The first step involves the geologist making a prediction based on historic statistics
and analogues of other basins and plays with similar geological characteristics, of the
chance of there being any hydrocarbons in the prospect. Some practitioners define
this chance of success estimate to be Hgeological riskK &Simpson et al., 5666'.
Sensitivity analysis can be used here to identify the key reservoir parameters in this
case.
5. Assess the chance of success based on historic statistics and analogues of other basins and plays
with similar geological characteristics.
9. 3se sensitivity analysis to determine the critical reservoir parameters.
;. #onduct a probabilistic analysis of reserves using +onte #arlo techniques. If necessary, perform a
further sensitivity analysis here by altering the shapes of the probability distributions assigned to
the reservoir parameters and changing the nature of the dependencies between the variables.
<. A*traction from the probabilistic output of the reserves calculation of some deterministic samples
Jfor e*ample, p5>, p?> and p6> &high, mid, low cases'.
?. 3se sensitivity analysis to determine the critical economic parameters.
@. (erform probabilistic economic analysis for each deterministic reserve case using +onte #arlo
techniques. If necessary, perform a further sensitivity analysis here by altering the shapes of the
probability distributions assigned to the economic factors and changing the nature of the
dependencies between the variables.
7. 3sing influence diagrams draw the decision tree.
=. "or each reserve case, recombine the chance of success estimated in step 5 and the economic
values generated in step @, through a decision tree analysis to generate A+0s.
6. 3se option theory via decision tree analysis and assess the impact on the A+0.
"igure ?.59: # 9 Step approach to in,estment appraisal in the upstream oil and !as industry
/e*t, the geologist performs a probabilistic analysis of reserves using +onte #arlo
techniques. The following formula is used to generate the estimate of the volume of
hydrocarbons recoverable from an underground prospect:
$ecoverable reserves T gross rock volume V net payNgross pay V porosity V
hydrocarbon saturation V recovery efficiency V formation
volume factor,
556
where, gross rock volume &)$0' is the total volume of the HcontainerK mapped out by
the geologistsF netNgross is the proportion of the container that is reservoir rock &for
e*ample, sand' as opposed to non2reservoir rock &shale'F porosity is a measure of the
fluid storage space &or pores' in the reservoir rock, as opposed to sand grainsF
hydrocarbon saturation is the proportion of fluid in the pore spaces that is
hydrocarbons as opposed to waterF recovery efficiency is the proportion of
hydrocarbons in the reservoir that engineers can actually get outF and, formation
volume factor describes the change in volume of hydrocarbons as they flow from the
pressure and temperature of the subsurface to the surface &:ailey et al., in press'.
The geologists, based on limited data, draw probability distributions for each of these
variables. In an ideal world, the individual distributions would be entirely data driven
J based on data derived from many porosity measurements, for e*ample. :ut, in
practice, the data available are often minimal. The geologists will suggest the shape
of the curve that is consistent with the small amount of data available. )eologists
often, for instance, draw analogies between the porosity of the rocks being e*amined
and the porosity of rocks from a similar previously e*ploited area &:ailey et al., in
press'. As indicated above in section ?.<, the shape of the distributions to be used is a
contentious issue. The distributions can vary enormously and they will be chosen to
fit different circumstances. A triangular distribution, for instance, might be chosen for
porosity if the e*perts were confident that they knew the minimum, most likely and
ma*imum porosities. A lognormal distribution might seem most appropriate for )$0,
indicating that e*perts think that there is a slightly higher chance of very high values
than of very low. %nce each variable has been assigned a distribution type, any
dependencies between the parameters are modeled. Section ?.< e*plained that due to
the lack of prescription in the literature, this is another difficult task for the geologist.
)eologists usually presume some correlation between hydrocarbon saturation,
porosity and recovery efficiency. The +onte #arlo simulation is then run using, for
e*ample, #rystal :allY or ]riskY, and the end result is a new distribution curve of
the range of possible recoverable reserve si1es and the probability of any particular
one occurring. Some analysts refer to the range of possible recoverable reserves as
Hgeological uncertaintyK &Simpson et al., 5666'. A further sensitivity analysis can
then be carried out so that the key reservoir parameters in this case can be identified.
59>
"rom the resulting distribution, the geologist reads off values of the possible
recoverable reserves and the chances of their occurrence, for input to the economic
model. These values need to be representative of the whole distribution and so it is
good practice to use the p5>, p?> and p6> values since these represent the highest,
mid and lowest reserve cases.
Then the economists for each reserve case, with input from other specialists as
necessary, build the economic model. This involves generating Hmost likelyK
predictions of drilling, capital and operating and abandonment e*penditures,
production volumes, oil price and e*change rate. (robability distributions are then
assigned to each variable. The dependencies between any parameters are also
modelled. Section ?.< indicated that these tasks are particularly difficult because of
the lack of prescription in the literature. The +onte #arlo simulation is run, again
using ]riskY or #rystal :allY, and the result is a probability distribution of the
range of possible /(0s and the probability of any particular one occurring.
Sensitivity analysis can then be used to identify the key parameters in this case.
3sing influence diagrams as necessary, decision trees can then be drawn up for each
reserve case. The organisationBs decision2makers ought to be involved in this process.
This ensures that the analysts capture the decision2makers beliefs and preferences in
the analysis. #ombining the chance of success estimate generated in the second step
with the /(0 prediction for each reserve case, an A+0 for each reserve case can be
produced. %ption theory, which perhaps most easily applied using :uckleyBs &9>>>'
advanced decision tree, can then be used to allow analysts and decision2makers to
assess the impact on the A+0 of future events.
0ariations of the approach could be used for development decisions, any production
decisions and for the decision of when to abandon production and how to
decommission the facilities. "or e*ample, when organisations are considering
developing a field, the question of whether there are any hydrocarbons present is
omitted, since e*ploration and appraisal wells have already established their presence.
They focus instead on whether there are enough hydrocarbons present for the prospect
to be commercially viable. In the language of Simpson et al. &5666' and !atson
&566=', the organisation is now interested in Hcommercial riskK and Hcommercial
595
uncertaintyK as opposed to Hgeological riskK and Hgeological uncertaintyK &This will
be discussed further in Section @.9 of #hapter @'. As an asset proceeds through its life
from e*ploration, through development to production and, ultimately, to
abandonment, the relative risk and uncertainty associated with it decrease, though the
relative risk and uncertainty may increase. &"or e*ample, at a recent Society of
(etroleum Angineers seminar in Aberdeen, +ike #ooper and Steve :urford
demonstrated how the relative risk and uncertainty associated with the +urchison
field actually increased with time'. 3nfortunately, most of the decisions have to be
made near the beginning of the assetBs life when the risk and uncertainty are
particularly high. This makes it paramount that organisations use probabilistic
methods for the generation of both reserves and economic estimates.
Some companies use a combination of deterministic and probabilistic techniques. "or
e*ample, /angea and Dunt &5667' describe how +obil used such an approach for
reserve and resource evaluation prior to their merger with A**on. The authors, and
presumably the company, believed that both methods have valid 8ustification for
utilisation and that when they are used 8ointly, they can provide greater insights into
the recoverable hydrocarbon volumes and the probability of recovering those
volumes, than when they are used in isolation. .uring e*ploration, proved and
probable reserves &as defined by the !orld (etroleum #ongress, see Section 7.< of
#hapter 7' were calculated deterministically. A +onte #arlo simulation was then run
to establish the cumulative probability distribution for recoverable hydrocarbons.
This curve and the deterministic results were then utilised to determine the possible
volumes &Hgeological uncertaintyK' and associated confidence factors &Hgeological
riskK' for each of these categories. The mean value of this probabilistic curve &the
e*pected value' was the case used for the economic analysis. /o indication is given
in the /angea and DuntBs &5667' paper as to how +obil conducted their economic
analysis. Dowever, clearly by using only one reserves case to run the economic
analysis, the economic impact of the high and low reserve cases was ignored. As
fields went into production and new information became available, +obilBs analysts
generated new deterministic and probabilistic estimates of the proven and possible
reserves of each field. /ear the end of fieldsB lives, +obil believed that there is a little
uncertainty associated with the reservoir parameters or the si1e of the field.
#onsequently, the company discontinued using probabilistic analysis at this stage, and
599
the production volumes &and associated confidence factors' were calculated
deterministically.
!hilst it is certainly true that during production risks and uncertainties are
significantly reduced, they are by no means eliminated. "or e*ample, the oil price
prediction used in the economic models may prove inaccurate, as was the case with
the :rent field. The physical structures could fail. This occurred with the facilities
for "oinaven, :alder and Sleipner &!ood +acken1ie, 5666'. As indicated earlier,
there is also the possibility of phenomena occurring that are out with management
control so2called Hacts of godK. Spencer and +organ &566=' refer to such acts as
Htrain wrecksK. They define a Htrain wreckK to be an e*ceptional event that is not
accounted for in the analysis. In a mature field, an e*ample of a train wreck is the
reaction of )reenpeace to the decommissioning of the :rent Spar. There are also still
significant investment decisions to be made. "or e*ample, well intervention and side
track decisions. "or these and other production decisions, it is evident from the
e*amples above that companies ought to use decision analysis techniques with
probabilistic input acknowledging the risks and uncertainties that remain. :y
selecting a single value, +obil were ignoring other probable outcomes for each
pro8ect variable &data which are often vital to the investment decision as they pertain
to the risk and uncertainty of the pro8ect' implying a certainty which does not e*ist.
Spencer and +organ &566=' describe the application probabilistic techniques to
production forecasting using the choke model &figure ?.5;' in :(. This model
considers the reservoir, wells, facilities and e*port decisions as a system analogous to
a pipeline with various chokes restricting flow. Aach HchokeK is a probability
distribution of either production or efficiency. These individual distributions are then
combined by +onte #arlo simulation. It is usually assumed that all the distributions
are independent. The authors recognise that, in practice, this is often not the case and
they highlight the need for this issue to be addressed. 3sing probabilistic techniques
for production decisions e*plicitly recognises the inherent uncertainty in the input
parameters. The authors claim that using these methods has reduced the gap between
actual and predicted outcomes.
59;
"igure ?.5;: Choke model (source: Spencer and 2or!an, 199A
This section has provided an indication of the way the tools identified in this chapter
can be used together. Since their use together is resource2intensive, the approach that
is suggested in figure ?.59 would only be appropriate for investment decisions that
require HsignificantK capital e*penditure. This is a relative measure, for e*ample, a
small petroleum company might regard the investment needed to acquire seismic data
as HsignificantK &figure ?.5', whereas for a large company, the sums involved only
become HsignificantK when it is considering whether to develop the field &figure ?.5'
&Section @.; of #hapter @'. 0ariations of the approach summarised in figure ?.59
could also be used in other industries with a similar business environment to the oil
and gas industry, for e*ample, the pharmaceutical or aerospace industries. In these
businesses, the investment decisions are similar in scale to the oil industry, also
characterised by high risk and uncertainty and have a high initial investment without
the prospect of revenues for a significant period.
#ommercially available software packages can be used to assist the decision2maker
with some of these steps. "or e*ample, +erak produces various tools such as
.ecision TreeY, (ortfolioY and (AA(Y &(etroleum Aconomic Avaluation (ackage'
which uses #rystal :allY to perform +onte #arlo analysis. ./0 &.et /orske
0eritas' have developed a software tool, Aasy $iskY, for preference theory analysis.
Dowever, currently there is no single piece of software that allows the upstream
decision2maker to utilise all the tools in their toolkit. Through recently established
collaborative relationships, the ma8or players &#SI$% &#ommonwealth Scientific and
Industrial $esearch %rganisation' Australia, +erak, )affney, #line ^ Associates,
$eservoir
potential
!ell
capacity
"acilities
limits
A*port
system
59<
!ood +acken1ie and ./0' are now working together in an attempt to deliver to the
upstream investment decision2maker the definitive software tool.
8(; *O-*+U0IO-
This chapter has answered the first research question posed in #hapter 5 by presenting
the spectrum of techniques available to the industry for investment decision2making.
This is not intended to be a comprehensive study of the mathematics governing and
underpinning each technique. This is widely documented elsewhere. The aim here
was only to give an overview of the methods and indicate current theoretical
capability. $ecent studies suggest that current practice is some way behind this
potential. Dowever, this research has been limited and it is apparent there is a need
for a study that establishes common practice in upstream investment appraisal. The
following chapter addresses this issue.
59?
*hapter 0i=
*urrent practice in investment appraisal in the
upstream oil and gas industry
59@
:(3 I-,RODU*,IO-
There has been much research published on decision2making &for e*ample, "ord,
9>>>F )unnF 9>>>F Akenberg, 9>>>F +arkides, 5666F Darrison and (elleteir, 9>>>F
+ilne and #hanF 5666F /utt, 5666F :urke and +iller, 5666F (apadakis, 566=F .ean
and Sharfman, 566@F Guinn, 56=>F +int1berg et al., 567@F #yert and +arch, 56@;'.
The numerous qualitative studies that have been conducted are useful for providing
broad insights into the field of decision2making. Dowever, very few of these studies
have e*amined the use of decision analysis in investment decision2making. Several
have focussed on the e*istence of formalisation and rationality in decision2making
&for e*ample, (apadakis, 566=F .ean and Sharfman, 566@' but few have e*plicitly
e*amined the use, and usefulness, of decision analysis in investment appraisal
decision2making. "ewer again have considered cases where the decision situation is
characterised by a substantial initial investment, high &absolute' risk and uncertainty
throughout the life of the asset and a long payback period, features that are common
in, though not unique to, the petroleum industry. Typically, where such research has
been undertaken, it has been conducted within one company, usually by an employee
of that organisation and has often not been published due to commercial sensitivity
&for e*ample, :urnside, 566='. There has only been one previous qualitative study
researching the use of decision analysis across the whole oil industry &"letcher and
.romgoole, 566@'. Dowever, as stated in Section ;.< of #hapter ;, this study focused
on the perceptions and beliefs of, and techniques used by, one functional area within
the organisations active in the upstream. Dence, its findings can only be regarded as
indicative rather than conclusive. There are also many quantitative studies of
decision2making. As indicated in #hapter 9, where these have been centred on the use
of decision analysis in investment appraisal decision2making by organisations, they
have only provided an indication of how widely used a particular decision analysis
technique is &for e*ample see studies by Arnold and Dat1opoulous, 5666F #arr and
Tomkins, 566=F Schuyler, 5667F :uckley et al., 566@ "letcher and .romgoole, 566@F
Shao and Shao, 566;F 4im, "arragher and #rick, 56=<F Stanley and :lock, 56=;F
!icks 4elly and (hilippatos, 56=9F :avishi, 56=5F %blak and Delm, 56=> and
Stonehill and /athanson, 56@='. They do not provide any insights, based on
behavioural decision theory, into the reasons why some techniques fail to be
597
implemented and others succeed, and, more importantly, which techniques perform
better than others do &#lemen, 5666'.
The research presented in this chapter differs from these studies. 3sing a qualitative
methodology, it attempts to integrate perspectives from individuals employed in a
variety of functions within organisations who are involved throughout the investment
appraisal decision2making process. This allows insights to be gained into issues such
as why organisations use certain techniques and yet re8ect others. #ombining this
with the review of the relevant behavioural decision theory literature &summarised in
Section 9.< of #hapter 9', allows the second research question that was proposed in
#hapter 5, which aimed to ascertain which decision analysis techniques upstream
companies use and to understand how they use them, to be answered.
The chapter first establishes which techniques are currently used for investment
appraisal in the upstream. $esearch by Schuyler &5667' and "letcher and .romgoole
&566@' has suggested that there is a significant gap between practice and capability in
the techniques used for investment appraisal in the upstream oil and gas industry.
#hapter ? presented the decision analysis tools currently available to the industry.
Some of these techniques have only been applied to the oil industry recently and
hence, were not available to companies at the time of these previous studies. The
chapter begins by drawing on the research interviews to establish first, which
techniques upstream companies now use for investment appraisal and second, if there
is still a gap between current theory and practice in investment appraisal decision2
making. This indication of current practice will be used in #hapter 7 to produce a
ranking of the companies according to the sophistication of the decision analysis tools
they use for decision2making. The chapter concludes by developing a model of
current practice in investment appraisal in the upstream. If there is a gap between
current practice and capability, this model will allow possible reasons for its e*istence
to be e*plored.
:(% ,) U0) OF D)*I0IO- A-A+C0I0 1C ORGA-I0A,IO-0
.rawing on the research interviews, this section establishes first the e*tent to which
companies are aware of and second, the amount to which they use each of the
59=
techniques identified in #hapter ?. This picture of current practice can then be
compared with the 6-step approach presented in figure ?.59 of Section ?.7 in #hapter
? that represented current capability.
The concepts of decision tree analysis and A+0
Awareness in the industry of the concepts of A+0 and decision tree analysis is high
and, in all but one of the companies interviewed, their use in investment appraisal
decision2making is commonplace. #onfirming the literature that was reviewed in
Section ?.9 of #hapter ?, the value of a decision tree is appreciated almost universally
in the upstream. +ost of the companies have been using decision trees for some time
and find the tool useful. Several respondents believe that decision trees are more
effective in organisational investment decision2making than techniques such as +onte
#arlo simulation because they encourage the e*plicit consideration of all the potential
outcomes of a decision. This, interviewees feel, is especially valuable when an
investment decision is particularly comple*. Some organisations have software
packages to assist with structuring and presenting their decision trees. The most
commonly used package is .ecision TreeY &produced by +erak'. The ma8ority,
however, are of the opinion that it is easier to draw decision trees by hand:
Hand then they say, H#an you put this decision tree into a drawing programI
And you go, HAhIK :ecause it asks for your hierarchies, sub2hierarchies or
whatever. And with our decision tree program thereBs an awful lot of
language.K &#'
/one of the companies reported using influence diagrams to structure their decision
trees. (earson2Tukey appro*imations are not employed by any of the companies in
decision tree analysis. .ecision trees tend to be used for all the investment decisions
throughout the life of an asset &see figure ?.5 and Section ?.9 of #hapter ? outlines
these decisions'. Dowever, in most organisations decision trees are not presented to,
or used by, the main board. This is issue receives further attention in section @.;.
$ecognising the folly of reliance on only one decision2making criterion &Atrill, 9>>>'
and echoing earlier observations by Schuyler &5667' and others &Arnold and
Dat1opoulous, 5666F #arr and Tomkins, 566=F Schuyler, 5667F :uckley et al., 566@
596
"letcher and .romgoole, 566@F Shao and Shao, 566;F 4im, "arragher and #rick,
56=<F Stanley and :lock, 56=;F !icks 4elly and (hilippatos, 56=9F :avishi, 56=5F
%blak and Delm, 56=> and Stonehill and /athanson, 56@=', companies report that for
significant decisions, a range of decision2making criteria is generated and presented to
the board. %rganisations weight these measures according to environmental
conditions and the particular decision under analysis. As several of the respondents
e*plain, an A+0 only tends to be generated by organisations when they are trying to
decide whether or not to drill an e*ploration prospect:
HThey are used for different things. /(0 is very important. A+0 is used on a
drill or donBt drill decision. It is used because the managing director likes it.
$%$ L$ate of $eturnM is used as well. !e always quote /(0 and $%$ in any
conversation. :ut the others are used.K &,'
and,
H.rill or not drill is A+0. $%$ is important we have a threshold J if an A^(
LA*ploration and (roductionM pro8ect doesnBt have a $%$ greater than a
particular threshold and we use /(0 to give us an idea of the si1e of the
pro8ect value.K &"'
%ne of the representatives from company $ e*plains why an A+0 is only calculated
for drilling decisions:
Hreally because once youBve made the decision to spend the money then
basically A+0 becomes a slightly meaningless term because A+0 tends to
be used more when you are risking an e*ploration prospect but once youBve
found something and you feel as though there is a good chance that you are
going to make money out of it, then really how do you manage that riskI So
itBs the sensitivity around the core, the base value. I would say A+0 would
tend to be used where youBve got significant levels of risk of failure, where
you are probably more likely to fail than to succeed and that would typically
be in an e*ploration venture. /(0 would definitely be used when youBve
found something and you are going ahead.K &$5'
$ecall from Section ?.9 of #hapter ?, the A+0 of an outcome is defined by
/ewendorp &566@' to be the product that is obtained by multiplying the chance &or
probability' that the outcome will occur and the conditional value &or worth' that is
received if the outcome occurs. The A+0 of a decision alternative is then the
algebraic sum of the e*pected values of each possible outcome that could occur if the
decision alternative is accepted. After the decision has been made to drill an
5;>
e*ploration well in a field and the presence of hydrocarbons has been confirmed, the
chance &or Hgeological riskK J for a full discussion see Section ?.7 of #hapter ?' of
there being a dry hole is 1ero and, consequently, the A+0 of the outcome Hdry holeK
is 1ero. Dowever, the dry hole is only one of the outcomes on the decision tree.
There are still many outcomes that could occur. "or e*ample, the field could contain
fifty million barrels or it might only contain ten million barrels. Aach outcome has a
chance of occurrence and a conditional value that would be received if the outcome
occurred. Dence, theoretically, at least, the A+0 could be used for all the decisions in
the life of an asset. As indicated in Section ?.7 of #hapter ?, this observation has led
some companies to differentiate between geological risk and uncertainty and
commercial risk and uncertainty. Such organisations define geological risk to be the
chance of there being any hydrocarbons and they perceive geological uncertainty to
be the range of possible volume outcomes given there is some hydrocarbons. An
A+0 can then be calculated for the decision to drill. %nce the presence of
hydrocarbons has been detected the focus then shifts to commercial risk and
uncertainty. #ommercial risk is defined to be the chance of the field producing
enough hydrocarbons to be commercially viable in the current and future economic
climates. #ommercial uncertainty is defined to be, given that the field is
commercially viable, the possible range of outcomes. An A+0 can be calculated
again at this stage.
As the following interviewee indicates, decision2making criteria appear to move in
and out of favour with management:
HL!eM use all of these. This one LA+0M is used most heavily here in
e*ploration than anywhere else. Ces we use all these. /(0 is the one that
we pay most attention to. $%$ gets people e*cited from time to time. They
8ust sort of disappear and come back again a couple of years later.K &/5'
As stated in Section ?.9 of #hapter ?, the A+0 of a decision alternative is interpreted
to mean the average monetary profit per decision that would be realised if the
decision2maker accepted the alternative over a series of repeated trials. The A+0
decision rule then holds that provided the decision2maker consistently selects the
alternative that has the highest positive A+0, then the total net gain from all decisions
will be higher than the gain realised from any alternative strategy for selecting
5;5
decisions under uncertainty. If decision2makers in the upstream vary the decision2
making criterion they use to choose e*ploration prospects, then they are failing to
satisfy the repeated trial condition of the A+0 decision rule. This occurs in some
organisations because there is a misunderstanding at board level of what A+0 really
means as the following respondent illustrates:
HLThere is a lackM of understanding of what A+0 means. (eople look at an
A+0 and think that is the value of the prospect not recognising that it is the
aggregated e*pected value of the various outcomes. IBve got a classic one
here. I showed to the board a portfolio of 6 ma8or pro8ects. All of which had
their own risk and uncertainty and theyBre highly related in that one or two of
them controlled whether or not others of them would go ahead. So if one of
them didnBt go ahead, for instance, there was a little satellite that wouldnBt go
ahead too. So you put it all together in a decision tree, you roll it all up and
you calculate an A+0 and associated with that A+0 there are other things
like e*pected #A(AW. And they looked at it and said Hmmm not very good is
it. It means our company is only worth S?> million.K And you say, H/ah,
youBre wrong. !hat we are saying is if you did them all, you could e*pect at
the end of the day some failures and some good ones and thatBs your value. If
you get clever and do the good ones first, you may already find that youBve got
S5>> million in your pocket and the clever thing to do is then to stop gambling
and go somewhere else $ightIK And again I was always taught positive
A+0 no matter the magnitude itBs good newsI :ut they were looking at the
magnitude as being an indicator of success.K &.'
In other organisations, as the following interviewee e*plains, it is not
misunderstanding of A+0 but the lack of multiple prospects that makes the A+0
decision rule impossible to adhere to:
HI mean for prospect analysis I think it is pretty standard to use an A+0
approach which is essentially the weakness in the A+0 approach, any
decision tree approach, is that the value that comes out maybe actually a value
that will never actually occur in practice. And if you take a very simple
approach to prospect analysis which I think a lot of companies do. This two
outcome model either a dry hole or a success of a certain si1e. Then what you
say is either a' I am going to lose R5> million or b' IBm going to make R5>>
million. The A+0 will come out at letBs say R9> million or R@ million. :ut
thatBs not going to occur. Cou are either going to lose 5> or make 5>>. So
how can that represent that decisionI %f course if you talk to people like
/ewendorp who published on this years ago, he would say of course itBs a
nonsense you canBt use it for that. Cou can use it as a comparative tool. If
youBve got a lot of prospects, youBve got a statistical database and over time it
will come out. CouBll achieve A+0. Then I come back to this problem
what if I only have one prospect, I donBt have a statistical sample here that I
can play with, how can I value itI !ell I canBt think of another way of doing
it. So it is a way of doing it but I have never been entirely happy with it.K &G'
5;9
(erhaps these observations help to e*plain why in this study, like SchuylerBs &5667'
earlier work, so many of the respondents reported difficulty in knowing whether a
pro8ect would be approved or not. +any of the professionals interviewed admit to
being confused about their companyBs decision policy. +ost are unsure about how
their company policy might make trade2offs among different decision criteria:
HI donBt know which ones are used for which decisions.K &"'
Indeed one respondent commented:
HLThe approachM changes annually, monthly, daily and also vertically, often
with a change in chief e*ecutive.K &$9'
The misunderstandings of decision2makers and the process by which they actually
make investment decisions are issues that will be discussed further in section @.;. The
focus in this section is on those decision analysis techniques that companies choose to
use for investment appraisal. In this regard, the level of awareness and usage of
+onte #arlo by upstream organisations will now be discussed.
$isk analysis using +onte #arlo simulation
Awareness of +onte #arlo simulation in the upstream is high. All but one of the
respondents recognised the technique and it is widely used to generate estimates of
prospect reserves. "rom the resulting probability distribution of recoverable reserves,
organisations typically select only one reserve case, usually the p?> or mean value, to
run their economic models on. This means that companies are ignoring the economic
impact of the high and low reserve cases. 0ery few of the organisations use +onte
#arlo at the prospect economics level. The reason companies choose not to employ
+onte #arlo to generate economic estimates is well described by this respondent:
H/o we donBt +onte #arlo our economics. And I had a discussion 8ust the
other day about whether we should be using +onte #arlo on our economics.
:ut in the amount of work involved in getting the input data for the economics
you know to get all the costings, and this sort of thing. ItBs hard enough to
get the data together to do an economic run based on the most likely or the
mean reserves estimate. That to get enough data and the right data to do it
probabilistically 2 we 8ust couldnBt do it. The system would break down. Cou
5;;
know we are over worked as it is. L!eM should be Lusing probabilistic
economicsM.
I suspect that no oneBs doing it for the same reason as weBre not, because of the
amount of work involved is so much greater than the amount of work involved
in 8ust, you know, 8ust single figure input economics. ,ust getting all the
sources together. Cou know weBre always up against a time pressure. It
always has to be now.K &,'
#onfirming earlier indications by Schuyler &5667', none of the sampled companies
routinely use +onte #arlo decision2making at the production phase of field
development &figure ?.5 and Section ?.9 of #hapter ?'. All the organisations resort to
deterministic analysis for production decisions. /angea and Dunt &5667' argue that
companies are 8ustified in discontinuing probabilistic analysis during production
decision2making since there is little uncertainty associated with the reservoir
parameters or the si1e of the field at this stage. Dowever, as indicated in Section ?.7
of #hapter ?, there are cases where the relative uncertainty has actually increased with
field life. +oreover, whilst typically the absolute uncertainty decreases with field life,
the relative uncertainty associated with, for e*ample, well2intervention decisions, is
significant.
As indicated in Section ?.< of #hapter ?, there are a number of theoretical limitations
of +onte #arlo simulationF the most significant of which are the lack of prescription
in the literature concerning the shape of probability distribution to be used to
represent the reservoir parameters of reservoir rocks of similar lithology and water
depth and the dependency to be used to represent the relationships between the
reservoir parameters. +ost of the organisations interviewed cope with this gap by
leaving the type of distribution and nature of the dependencies used to the discretion
of the geologist. )eologists report that they decide the distribution shape and
dependencies based on a blend of intuition &:aumard, 5666 p@7', tacit knowledge
&(olyani, 56@@' and 8udgement.
$espondents are divided on whether varying the shapes of these distributions affects
the output and, correspondingly, whether there is potential for the non2discerning to
manipulate the results. "or e*ample according to some respondents:
5;<
HThere are a few unscrupulous people that cheat like cra1y. The big one
that is abused beyond belief is dependencies. And the programs are, I donBt
know, lacking in robustness. The one that weBve got, you know, is a classic
one, porosity versus water saturation. ThereBs normally if you plot the data a
correlation, but when does a correlation actually suggest a dependency of less
than one and is it >.?,>.@,>.7I IBve actually, way back in my youth, when I
was mucking about with all this, actually shown, that without violating
anything, you could quite easily alter the Lrecoverable reservesM by 9>E. This
is fundamental to the crooked. Averyone here is in the business of procuring
funds for their pro8ects. ItBs not a question of is it right or wrong. ItBs because
IBve done this work and it suggests to me that this is a 8olly good pro8ect. /ow
the man across the corridor is competing for the same funds so it is a
competition. And may the best man win and nobody sets out to cheat but K
&#'
and,
Hpeople with a better understanding of statistics were able to HscoogleK and
skew the outcome by putting in a particular distribution shape so you donBt
actually change the numbers you 8ust change the distribution and that can
change the output.K &.'
!hereas others argue:
Hthe shapes of the distributions is relatively insensitive thing.K &/',
Hthe type of distribution you use is not that importantK &$5'
and,
Hit seems to be quite robust to any type of distribution that you put inK &$<'
In an attempt to remove discretion from the analyst and impose more rigour on the
process, some organisations do prescribe which distribution shape is used for each
reservoir parameters in a +onte #arlo simulation:
H!e have recommendation that we use beta distributions. And thatBs for
consistency because we tried calculating the reserves of a prospect using the
same input data with different distributions and we got quite a range of
numbers out. So for consistency use beta and thatBs it.K &,'
/ewendorp &566@ p;=7' warns of the dangers of using distributions in this way.
%ther respondents also warned against such this practice, arguing that it HforcedK the
data. These interviewees believed that distribution shapes should left to the discretion
of the analyst so they can be Hdata2ledK:
5;?
HSome companies will deliberately impose a lognormal distribution on
everything they do. LThis isM based on the belief that all of these ranges are
lognormal. I strongly disagree with that. I think itBs invalid and incorrect
to do that and LthatM you should be guided by the data.K &)'
Dowever, questions have been raised over which data companies should be led by
&Simpson et al., 5666'. Snow et al &566@' argue that statistical analysis of parameters
from nearby wells is a valid method of determining the shape of input distribution to
be used in a +onte #arlo analysis. Dowever, petroleum reservoirs are heterogeneous,
and reservoir parameters vary from sample to sample. Therefore, Simpson et al.
&5666' argue, reservoir modelling requires field2wide weighted average values,
derived from detailed mapping of parameters, should be used to derive these
probability distributions.
All respondents agreed that the lack of prescription in the literature contributes to the
overall dissatisfaction with the process and to companiesB reluctance to endorse +onte
#arlo simulation:
Hthat is actually one of the reasons why some people are uncomfortable with
+onte #arlo simulation because they are not convinced that dependencies are
properly handled. They are suspicious of a mathematical black bo*. And
thereBs a relationship between porosity and water saturation for e*ample, they
are not convinced that that it is recognised. Aven if you put in a porosity
distribution and a porosity water function they are still uncomfortable.K &('
There was broad agreement that a study indicating the shape of distributions and the
nature of the dependencies that should be used for different reservoir parameters, in
different geological formations at various depths, is long overdue:
HI wish someone would come up with a :ritish standard for these things J it
would make life a lot easier.K &/5'
and,
HLTheM ideal scenario is that there would be an industry standardK &$<'
All of the companies use some software to assist with the +onte #arlo simulation.
The most popular packages are #rystal :allY, ]riskY and (AA(Y. There was a
general recognition that whilst the mechanics of the simulation is straightforward:
5;@
Hthe clever bit is in the process that goes on before you press the button and
the numbers are churning round in LtheM +onte #arlo LsimulationM. The clever
bit is in the model that you set up where youBve got the risk and youBve got
the relationship. ThatBs the clever bit. So you can have a fantastic tool that
does +onte #arlo inside and out but LitBsMgarbage in2garbage out.K &/9'
The respondent from company . also stressed:
Hlike a lot of black bo*es youBve got to be careful that you understand the
input.K &.'
In #hapter ?, three other techniques were highlighted as being useful to the oil
industry. (reference theory has been applied to oil industry investment decisions in
the literature since the 56@>s. Dowever, software has only recently become available
to assist with the generation of individualsB preference curves. %ption and portfolio
theories are tools from the finance industry that have only recently been adapted to
petroleum investment decisions. #onsequently, at the time of the previous studies
into the use of decision analysis techniques by the industry &for e*ample, Schuyler
&5667' and "letcher and .romgoole &566@'', these tools were not widely perceived to
be particularly applicable to the oil industry. Dence, the findings from this research
concerning the levels of awareness and usage of these tools in the upstream are
particularly interesting.
(ortfolio theory
Awareness of portfolio theory in the upstream is low and its usage is even lower.
%nly three of the organisations interviewed fully endorse the use of portfolio theory.
Dowever, most companies had an intuitive grasp on its fundamental principles. This
is well illustrated by this quote where the interviewee unwittingly describes the
difference between diversifiable and non2diversifiable risk:
H(ortfolio Theory .o we do thatI /ot as such. To a certain e*tent. I mean
one of the things at the moment is with the oil price being so low and gas
prices perhaps still holding up thereBs a shift from oil to gas in the portfolio.
So on that level yes. .o we look at individual pro8ects and say, Humm thatBs
risky, better have a safe oneIK /o. I would say not. Cou can get rid of some
of the risks but ones like oil price, e*change rate much less likely to be able to
mitigate those.K &/9'
5;7
As such, some respondents reported that while prospects are not analysed according
to the rigors of portfolio theory, before drilling a prospect, it is HscreenedK to see if it
fits with various organisational criteria. "or e*ample, some companies will only
operate in areas where there is low political risk whereas others prefer only to e*plore
where they know they will be the only operators. The following two quotes are
indicative of current practice:
HThe starting point, if you like, is that we have identified and review
periodically, so2called core areas in particular core countries. There are certain
areas of the world, for the sake of argument, South Aast Asia which we have
elected to not invest ourselves in because we are fully occupied elsewhere and
we think get better returns elsewhere. If we want to enter a new country one
of our responsibilities in LtheM commercial LdepartmentM is to maintain what we
call country evaluations. So within our areas of interest, we keep more or less
current evaluations of countries from the point of view of political, economic
stability, working environment etc. So the first point of call if the
e*plorationist want to go for the sake of argument Athiopia, then we would
need to consider the overall climate in the country and the technical
prospectivity and combining those two we then put a broad brush proposal to
our #hief A*ecutive. And this is going into a new country because he has to
sign off on any venture of that sort.K &(',
and,
H!e at this moment, I can only speak for what we are doing here in the 3.4.,
there has been a sort of strategic decision made beforehand about where we
should operate and how we should operate and so within that framework is
where we are now currently working. ThatBs mainly in the southern gas base.
So that is where we stand. So yes. There are strategic decisions and we sort
of try to test the waters every now and then to see what head office feels about
us going into certain directions.K &)'
In the literature, some authors &Simpson et al., 9>>>' argue that portfolio theory is
particularly applicable to small companies. Dowever, in this study those smaller
companies that were aware of the technique had re8ected its implementation because
they believed they had insufficient properties to constitute a HportfolioK:
H!e tend not to be spoilt for choice for investment opportunities. !e donBt
tend to need to rank development opportunities either in terms of risk or
reward because we have pretty much only got one or two going at any one
time. ...(ortfolio theory would certainly be more valuable in a bigger company
than ours.K &/5'
5;=
%ption theory
0ery few of the respondents were aware of real option theory, and in all cases of
awareness, the interest had not translated into use. The companies reported finding
the technique very complicated and the theory difficult to grasp. This comment from
one respondent was typical:
H%ption theory weBve been getting, not me personally, but people have been
getting e*cited about option theory but I think itBs run a bit out of puff a little
bit here at the moment. There are some particular advocates here but nobody
has been able to demonstrate it at least here, that on the ground and in practice,
it is very helpful. !hether thatBs right or not, I donBt knowK &/5'
(reference theory
%nly four of the respondents were aware of preference theory and none of them
reported using the technique as part of their investment appraisal decision2making
process. The ma8ority was of the opinion that it would be:
Hdifficult to convince the hard2nosed asset manager that they should use
such a process.K &/5'
3sually their level of knowledge of the technique was based on attending a workshop
or seminar by consultants where the technique had been reviewed.
The reason that option and preference theories and, to a lesser e*tent, portfolio theory,
are so rarely used by organisations is e*plained by one of the representatives of
company /:
HThereBs a lot of interesting things at the conceptual level but when it
comes down to standing in front of the directors and trying to help them
make a better decision regarding an issue, thereBs a subset I think of these
tools that are useful in doing that. +onte #arlo and decision trees are
about as far as it goes here.K &/5'
The observations in this section have indicated which techniques are being used by
organisations in their investment appraisal decision2making. "or e*ploration
decisions, most companies use +onte #arlo simulation to generate estimates of
5;6
prospect reserves. They then run their economic models on only one reserve case.
Typically, +onte #arlo simulation is not used for economic analysis. In production
decision2making, the ma8ority of companies only use deterministic analysis. %ption,
portfolio and preference theories are hardly used at all by any firm. #omparing this
approach with the 6-step approach outlined in figure ?.59 &Section ?.7 of #hapter ?',
this study has clearly confirmed suggestions from the earlier empirical research, and
established, unequivocally, that there is a gap between current theory and current
practice in the quantitative techniques used in investment appraisal in the upstream oil
and gas industry.
The following section builds on the discussion above. It draws on the interview data
and the behavioural decision theory literature &summarised in Section 9.< of #hapter
9' to gain insights into how organisations use their decision analysis tools and how the
decision2makers use the results from the analysis to make decisions. "rom this, it is
possible to suggest why there is a gap between current practice and capability in
investment appraisal in the upstream. A model of current practice in investment
appraisal in the upstream oil and gas industry can also be developed. This model is
presented in section @.<.
:(5 ,) I-?)0,M)-, A''RAI0A+ D)*I0IO--MA.I-G 'RO*)00
#onfirming SchuylerBs 5667 study, the findings from the research presented in this
thesis indicate that decision analysis techniques are being introduced slowly into
upstream organisations. .espite the application of decision analysis techniques to the
oil industry in the literature in the 56@>s &)rayson, 56@>', the ma8ority of upstream
representatives report that their organisations only began using them within the last
five years. $espondents typically e*plained this trend in two ways. "irstly, several
claimed that previously the computing power was insufficient to allow the use of
decision analysis techniques to be automated and hence their company had decided
against their implementation. Secondly, others perceived that the increasing risk and
uncertainty in the operating environment, as discussed in #hapter ;, had contributed
to their organisationBs recent interest in decision analysis. +ost companies first use
decision analysis tools on particular fields before recommending employing them
company2wide on all prospects and fields. This is confirmed by the tendency for
5<>
organisations to publish in industry 8ournals, such as the Gournal of %etroleum
)echnolo!y, accounts of using decision analysis techniques on specific cases &for
e*ample, Spencer and +organ, 566='. The ma8ority of the sampled companies have a
cost threshold that they use to indicate those decisions to which decision analysis
techniques ought to be applied. $eflecting their different attitudes to risk, in the
smaller companies this value is lower than in the larger organisations. Therefore,
decision analysis is used on a higher percentage of the decisions in small
organisations than in larger companies.
+ost companies have not altered which decision analysis techniques they use or how
they use them, since they first introduced the techniques. In some cases, corporate
adoption of the tools was accompanied by the production of manuals, which outlined
their new approach to investment appraisal, the introduction of corporate definitions
of risk and uncertainty and the instigation of training programs for staff. Such
organisations are reluctant to change their approach and be forced to repeat this
process. This means that in some companies, even though they are aware their
approach is not as sophisticated as it might be, they continue to use it:
HCes. IBm recommending changes to it. ...IBve got an alternate system that we
could go to. The problem is that the company only went to this process,
from having nothing really at all, several years ago, so they are loath to change
it again. And thatBs the problem. !e are locked into a system thatBs
inadequate and theyBre loath to change it to anything else. And thatBs cra1y.K
&:'
In other companies, the reasons for the failure to update the techniques they use or to
modify how they use decision analysis, are endemic within the organisation. This
section aims first to identify these reasons and, second, to understand their sources.
This will allow the author to e*plain why there is a gap between decision analysis
theory and its use in practice.
In Section 9.9 of #hapter 9 decision analysis was defined to be a normative discipline
within decision theory consisting of various techniques and concepts that provide a
comprehensive way to evaluate and compare the degree of risk and uncertainty
associated with investment choices. In addition, in this section of #hapter 9 literature
was highlighted that indicated that the definition of risk and uncertainty that the
5<5
decision2maker adopts affects the method that they use to cope with the risk and
uncertainty &-ipshit1 and Strauss, 5667F :utler, 5665F )randori, 56=<F Thompson,
56@7'. In some of the upstream companies interviewed, the organisation had no
corporate definition of risk and uncertainty. "or e*ample:
HI donBt know what youBve found in other companies, but I would say that
thereBs about as many different definitions of risk and uncertainty in our
company, as you found in your literature search.K &)'F
HCes every time we start to discuss risk we have arguments and rows.K &.'F
H.ifferent people have their own definitions and their own way they would
like to look at it. So if I go in speak to someone about their definitions of risk
it depends on what asset team they are in. Trying to get consistency of
approach is difficult. Aven if you speak to people with the same 8ob title
within the asset theyBve got different definitions.K &/9'F
and,
H:ut I do say that when you talk about risk, I think thereBs quarters here where
you would hear folk say itBs fundamental to do a risk assessment. :ut
normally itBs a risk assessment of health, safety and environmental. Are we
going to kill anybodyI Are we going to damage the ecosystemI Are we going
to pollute the environmentI .ifferent types of risk. Again itBs a
misconception. !hen I saw the risk analysis manual here, I was in seventh
heaven, but it was upsetting the breeding patterns of fish or something ...
!ithin any company, within any department, within any team, thereBs different
definitions.K &#'
In most of these companies, only very basic decision analysis techniques were used
&company / is the e*ception. In this company, the respondents described how it is
widely recognised that there are multiple definitions of risk and uncertainty within
their organisation. !hen employees communicate about risk and uncertainty they are
e*plicit about their definitions and perceptions of the terms'. In other companies
where e*plicit corporate definitions of risk and uncertainty have been introduced, the
organisation typically used more decision analysis techniques and had a more
formalised investment appraisal process. #learly then, organisations use of decision
analysis is affected by the corporate perception of risk and uncertainty which, in turn,
is a product of the decision2makersB beliefs. The sources of decision2makerBs beliefs
will now be e*amined.
5<9
As indicated above, decision analysis has only been introduced into most
organisations within the last five years and, consequently, most of the current chief
e*ecutive officers &#A%s' of organisations have often not been introduced to its
concepts throughout their careers. This point is well articulated by one respondent:
HI think there is a definite age imprint on decision2making. TodayBs #A%s
tend to be in their fifties now, and grew up corporately in the 56@>s when slide
rules and log tables were the norm. The young guns are much more
comfortable using Ldecision analysisM but corporately have to still climb to the
highest level.K &S;'
This situation is e*acerbated since even when companies choose to use decision
analysis techniques and believe that their introduction requires staff training, the
training is often only given to technical staff:
H!e run an uncertainty workshop which is part of the compulsory training
programme for new, mainly subsurface, staff and thatBs a four day long
workshop where we look at some of the statistics and theory and we go
through a whole series of worked e*amples. :ut itBs now being pushed at all
of the Hchallenge graduatesK, all the people 8oining the company, it is part of
their core skills. "or the people who have 8oined the company within the last
three years itBs fine because they are going through that. ItBs more of a
problem for the people who have been in the company say five or ten years.
ThatBs difficult.K &$5'
This lack of knowledge significantly affects top managementBs ability to understand
the philosophy of decision analysis. "or instance, there is a tendency for management
to prefer to communicate deterministically:
HThe reliance on one number is hard to get away from. It tends to go all the
way up. Aven LatM the highest level, even the managing director level, they
like to know, H!ell, whatBs the numberIK Aven at board level, they donBt tend
to deal with numbers for the ranges.K &$5'.
This preference for deterministic analysis then permeates the entire organisation:
HI still get the reaction if you ask people If you go to a cost engineer, the
die hard cost engineer, that has been in the shipyard all his life, and you go to
him and say, H!hat IBm after is how bad it could be and how good it could
be.K The usual story. H!hat do you mean how good it could beI This is what
IBm telling you itBs going to be.K Cou know itBs going to be S?> million and
the sheer concept that it could be anything different from that number heBs
given you is completely alien and at that end what theyBll say is, HCou want a
5<;
range on thatI !ell, itBs plus or minus ten per cent,K which is completely
pointless. So, Lit isM still a problem.K &/5'F
HAngineers like to deal with units and this is the number.K &$9'F
HThe biggest problem weBve got is that fact that we are deterministic. !eBve
always got to have some case to build action.K &.'F
and,
HIBm sure this must be a consistent observation. !e do all this beautiful
simulation of the distributions but people still want one figure. Cou could say
to them, HThe range is this, or your e*pectation is this at various probability
levels, youBve got say <>E chance of finding this and then you know ;>E
chance of finding this larger figure and so on.K Cou know a nice little
cumulative distribution. (eople donBt look at it. They want one number,
H!hatBs the meanI !hatBs the e*pected valueIK So sometimes I question
why we do it because people 8ust land on one number.K &D'.
"urthermore, several interviewees reported that their managers do not see any value in
using decision analysis. These respondents believed that this situation is e*acerbated
because there is no empirical study indicating that using decision analysis techniques
adds value to organisations. Some of the respondents also reported that the decisionJ
makers in their organisation do not understand the concept of decision analysis and
indeed perceive it to be a threat. This is well described by the contributors from
companies S and #:
H.ecision trees or A+0s, when viewed as traditional end of pro8ect
recommendations, show what decision should be made. There is no discretion
required. Dere you have a process that is providing the answer to the problemF
what is the role of the high level decision2makerF all can see what the answer
is J this is a very real threat to the decision2maker.K &S;'F
and,
H(eople think LhereM that by using LaM probabilistic approach you are actually
throwing out the essence of the businessK &#'
The manager interviewed at company / confirmed these observations:
H:lind faith is a better technique Lthan any decision analysis techniquesM
because then you are the boss.K &/9'
5<<
Such perceptions of decision analysis are closely aligned to the rational model
&Darrison, 566?', operational research &"rench, 56=6' and to the first definitions of
decision analysis proposed in the 56?>s and @>s &$aiffa, 56@='. These managers
believe that decision analysis is a purely normative tool, which removes discretion by
dictating choice. Such definitions do not emphasise the distinctive features of
decision analysis that distinguish it from the rational model. These misconceptions
affect the way decision analysis is used by organisations in two ways. Aach will be
discussed below.
The misconceptions cause divisions and communication difficulties in
organisations between those that understand the capabilities of decision analysis
and those that do not. %ften this is between those producing the analysis and the
decision2makers. This situation is e*acerbated and perpetuated since often in
companies the decision2makers are not involved in the process of generating the
analysis. They are often only presented with summary decision criteria. This has
three implications. "irstly, it means that managers are not educated sufficiently in
decision analysis to question the analysis. This could result in them accepting a
flawed pro8ect and which will subsequently fuel their distrust of decision analysis.
Secondly and more likely, it means that the decision2makers ignore, or reduce the
emphasis on, the analysis which, in turn, affects the motivation of employees to
compile the analysis and means that managers do not become educated in decision
analysis techniques. According to one respondent:
HItBs very interesting in those discussions how much importance Lis given to
the analysisM, because he Lthe decision2makerM doesnBt really grasp, I donBt
believe, whatBs going on here. So you know you do all that work and you go
to him, and it comes back down to, H!ell, what do you thinkIKand he has
his preferred advisors. So it comes down to sometimes what his preferred
advisors think who might not wholly understand what is going on down at the
probabilistic level either, or have not have been involved. So thereBs an awful
lot of input here from the people who have the trust of the leaderThe
decision2makers donBt get it. They go on opinion. They also go on the people
who they trust the best. That is very clear here./ow we still do this
Ldecision analysisM but it might not carry one bit of weight if people who are
the opinion holders if you like 2 the trustees, the most trusted employees 2 if
they donBt buy it.K &"'
5<?
The issue of trust is discussed in more detail below. The third effect of decision2
makers lack of involvement in generating the analysis is that it means that the
decision2makersB preferences, beliefs and 8udgements are not captured and
included in the analysis which must contribute to any inherent reluctance to accept
its recommendations.
Some companies have attempted to overcome these difficulties by introducing a
structured process for gaining management input to the decision2making process.
These companies are typically the larger organisations where employees are not
personally known to the decision2maker. This practice encourages communication
between analysts and decision2makers. #onsequently, it improves the efficiency of
the process in numerous ways, not least, by ensuring that the assumptions that the
analyst has underlying the analysis are consistent with the decision2makersB
opinions. This ought to result in fewer pro8ects being re8ected that reach the end of
the analysis process. It should also improve managersB understanding of, and
attitude toward, decision analysis.
In small companies, there are usually fewer opportunities and the levels of trust
tend to be higher since employees are usually known to the decision2maker. The
decision2makers are more naturally involved in the process and hence, generally,
companies do not think it necessary to have formal management Hbuy2inK to the
analysis process. This is well illustrated by one respondent:
HAnd the other thing I guess in our organisation is that we have direct
access to all decision2makers. I mean we LareM, in terms of people, really quite
small. I mean I can call up the president and #A%. DeBll call me if IBm the
person who can answer a particular thing. De wonBt go through the president
over here or the general manager of the department. DeBll 8ust give me a bu11.
De knows my e*tension and you know it might be, H!hat percentage interest
do we have in such and such a licenseIK or, H!hat do you think of thisIK or,
H.o you know anybody in such and such a companyIK The lines of
communication are so much easier. I mean heBs coming over here in a
couple of weeks and heBll come in and sit down and heBll make it very clear to
the individuals. LDeBll say,M H-ook this is what I want to see. I loved that
pro8ect you did before but IBm sorry I had to turn it down but, really, this is
what bothered me about it,K or, HIBm glad we did that and keep going and
bring me another one like that.K.Cou know we all speak the same kind of
language.K &A'
5<@
There appears to be a relationship between managementBs attitude toward decision
analysis and company culture. In companies where managers believe decision
analysis is valuable, the culture is Hnumbers2drivenK. In those organisations where
the decision2makers do not perceive decision analysis to be important, the
decision2making culture is Hopinion2drivenK. This will be labelled relationship one
here. This relationship is directly related to three other trends that are observable
in the interview data. These will be outlined here.
#ompare the following:
H!e are a numbers oriented company. The boss wants to see numbers
and he wants to see numbers 8ustified.K &/5'F
and,
H,oe :loggs down the corridor likes it a lot you know gives it @ out of
5> and we would like to do it. And thatBs the decisionK &A'.
Avidently then, there is a relationship between the use of decision analysis and the
culture of the organisation. In companies where the culture is Hnumbers2drivenK,
more decision analysis techniques tend to be used than in organisations with
Hopinion2drivenK cultures. This will be referred to here as relationship two.
In Hopinion2drivenK companies when decision analysis techniques are used, they
tend to be poorly implemented and supported. In companies that are Hnumbers2
drivenK, decision analysis techniques tend to be well supported and their use
encouraged. This will be referred to as relationship three.
"urthermore, there is a relationship between the formalisation &note, formalisation
does not imply sophistication' of the analysis and the level of employee
satisfaction with the process. Typically, in those companies where the procedure
for using decision analysis techniques is well defined, respondents generally felt
the analysis worked well. In others, where the process is less well defined, the
analysis often has numerous gaps and, generally, levels of dissatisfaction with the
approach are high. In these companies, the analysis process is often gone through
5<7
only to satisfy bureaucratic procedures. This will be labelled relationship four. In
larger companies, there is more of a need for formalisation for evaluating prospects
firstly to allow the relative ranking of opportunities and, secondly, because, as
-angley &566? p@<' noted:
Hthe more strategic power is shared among people who cannot quite trust
each other, the more formal analysis may become important.K
The author stresses that this observation does not imply that formal analysis should
be perceived as purely political tool in such cases, but simply that it has a dual role
in decision2making in large organisations:
H!hen used for gathering information, Lformal analysisM may help determine
and improve the substance of decisions directly, as most of the literature
indicates. :ut it can also help bind individualsB decisions together to create
organisational decisions through communication, direction and control, and
symbolism. The second, political role should not be automatically despised.
%n the contrary, when different organisation members do not necessarily have
the same goals or the same information sources, analysis helps to improve
decisions indirectly by ensuring that ideas are thoroughly debated and verified,
and that errors in proposals are detected before implementation.K &-angley,
566? p@<'.
+oreover, because of their si1e, such organisations are also more likely to have
individuals and departments using their own approaches. +ore management
involvement promotes consistency. To this end, some companies have also
introduced a peer review system. It is well described by this respondent:
HThe process we use for discussing risk is what we call peer review. This is a
shared learning e*ercise. !e will take the people who work the prospect and
essentially self2audit. ItBs not a process where management sits in 8udgement
because very often managers are generalists not specialists, especially in the
34. !e get the people who work the prospect, and a group of their peers
from within the organisation. $ecently we brought people from Australia and
)ulf of +e*ico to consider some pro8ects we were thinking of developing in
Angola. So, we actually spent a lot of money for that peer review process.
The people we brought in, we brought some technical e*perts for some
specific technologies and we brought some geoscientists who were working a
similar play in a different basin in the )ulf of +e*ico J so they have a great
knowledge to bring to bear on the situation and subsequent decision2making.
Then we went through risking session.K &4'
5<=
#ompanies vary significantly in the e*tent to which their decision2makers rely on
the analysis in making their final decision. !hilst most companies re"uire
decision analysis to be undertaken for investments above their cost threshold, the
e*tent to which the final decision is influenced by the data, appears to be
contingent on the four interdependent relationships outlined above.
In companies where managers are convinced about decision analysis, the culture is
Hnumbers2drivenK, the use of decision analysis is encouraged, formalised and well
supported, and employees are generally satisfied with their companiesB investment
appraisal process. Then the decision2maker relies on the results from the analysis
to make decisions. This is not to say that the decision is taken solely based on the
analysis. .ecision2making will always be ultimately an act of 8udgement.
Dowever, since the decision2maker has been involved in generating the analysis,
then its results are unlikely to contradict hisNher sub8ective 8udgement about the
particular investment opportunity. At the very least though, the analysis informs
the decision. If the analysis suggests that a pro8ect is not viable, and the decision2
maker still wants to go ahead, because of some bias or feeling that heNshe has not
been able to articulate and include in the analysis, they are doing so well informed
about the potential consequences.
In companies where managers are unconvinced about the value of decision
analysis, the company is largely Hopinion2drivenK and the use of decision analysis
is not formalised or encouraged. .ecisions in these companies are perceived by
the employees of the organisation to be influenced more by opinion and HfeelingK
than numerical analysis:
HAnd if you go through a structured decision process and you calculate an
A+0 and it is highly negative but your guts say this is a good thing to do
youBve then got two choices you can then go back and fiddle the numbers
or you can 8ust overrule the result and say that this is strategically good for
us.K &.'F
and,
HI have seen bidding strategy meetings held with senior management
where you would come forward with all of these Ldecision analysisM
evaluations and there the psychology in the meeting would override many
5<6
times the logic that had been developed using these probabilistic numbers.
Somebody likes something and suddenly the money would double. I saw that
many times.K &)'.
The observations above have been summarised in table @.5.
MA-AG)M)-, U-D)R0,A-D D)*I0IO-
A-A+C0I0
MA-AG)M)-, DO -O, U-D)R0,A-D
D)*I0IO- A-A+C0I0
The decision analysis approach used by the company
is formalised. %ften manuals are available to
employees. The manuals detail how the limitations and
gaps in the techniques &for e*ample, the distribution
shapes to be used in +onte #arlo simulation' are to be
overcome.
The decision analysis that is conducted is likely to
be lacking in definition, structure and
sophistication. Amployees are given no direction as
to how to deal with the limitations of the analysis
techniques.
.ecision analysis software available throughout the
%rganisation.
$estricted access to decision analysis software.
Amployees know the decision policy used by the
company.
Amployees do not know the decision policy used by
the company.
#onsistent definitions of risk and uncertainty. /o company definitions of risk and uncertainty.
.efinitions change within and between
organisational functions.
All employees have the ability to understand and
communicate probabilistically.
Amployees prefer to communicate deterministically.
)ood communication between the departments
compiling the analysis.
(oor communication between the departments
compiling the analysis.
+otivation to conduct analysis is high. +otivation to conduct analysis is low.
.ecision analysis perceived to be a useful tool for
quantifying risk and uncertainty.
.ecision analysis perceived, particularly by
management, to be a threat.
Aach prospect is sub8ected to peer review. There is no peer review system for prospect
evaluation.
.ecision analysis is part of the organisationBs culture. .ecision analysis is not part of the organisationBs
culture.
Amployees trust the results of the analysis. Amployees do not trust the results of the analysis.
Avery employee required to attend training in
decision analysis.
There is no training in decision analysis.
+anagement committed to decision analysis. +anagement not convinced by the value of decision
analysis.
+anagement involved in generating the analysis. Analysis conducted low down organisation.
+anagement only presented with decision2making
criteria.
+anagement likely to follow the decision alternative
suggested by the analysis.
+anagement less likely to follow the decision
alternative suggested by the analysis. They believe
their 8udgement in superior to the analysis.
Table @.5: 1r!anisations7 use of decision analysis
This table investigates the different ways decision analysis is used by organisations in
the upstream. It distinguishes between use of decision analysis where managers
understand decision analysis and the use of the techniques when managers do not. %f
5?>
the sampled companies, none e*hibit all of the attributes of either column. +ost are
placed somewhere on a continuum between the two e*tremes. The table clearly
highlights that decision2makersB attitudes towards decision analysis are one of the
main determinants of an organisationBs use of decision analysis techniques. As such,
the decision2makersB attitude toward decision analysis can be identified as one of the
factors that directly causes the gap between current theory and current practice in use
of decision analysis in investment appraisal decision2making by the upstream oil and
gas industry. This assertion is supported by a similar observation from 4unreuther
and Shoemaker &56=>':
H!hen decision theory analysis is viewed as a multi2stage model for rational
choice among alternative options, its impact on organisational theory and
managerial behaviour tends to be less than might have been hoped for or
e*pected.K &Thomas and Samson, 56=@ reproduced in "rench, 56=6 p577'
This section has provided an overview of the investment appraisal decision2making
process in the upstream. In particular it has highlighted that:
There is a relationship between the culture of the organisation and the decision2
makerBs perceptions of decision analysisF
there is a relationship between the use of decision analysis and organisational
cultureF
there is a relationship between the culture of the organisation and the e*tent to
which decision analysis is encouragedF
there is a relationship between the level of formalisation of the investment
appraisal process and employeesB satisfaction with the processF and
the actual decision &the HisK' generally, only deviates from that recommended by
decision analysis models &the HoughtK', when the analysis is poorly implemented,
unsophisticated and the decision2makers are unconvinced of the value of decision
analysis.
It has also suggested that at least part of the reason for the gap between practice and
capability is that:
5?5
There are theoretical gaps in some of the techniques &for e*ample, the lack of
prescription in +onte #arlo simulation of the shape of the probability distribution
to be used to model the reservoir parameters of reservoir rocks of similar lithology
and water depth'F
decision2makersB perceptions of decision analysis are closely aligned to the earlier
definitions of decision analysis and decision2makersB perceive decision analysis to
be a threatF and
decision2makers are not convinced of the value of decision analysis.
The section has also indicated that current situation is e*acerbated since there has
been no study conducted that has found a link between use of decision analysis and
good organisational performance.
:(7 A MOD)+ OF *URR)-, 'RA*,I*)
The observations e*pressed in the previous sections are captured here by a model of
current practice in investment appraisal in the upstream oil and gas industry. The
model presented has been modified and developed by abstracting from the insights
into decision2making at different levels within operating companies gained from the
research interviews discussed in sections @.9 and @.;, and is informed by the
behavioural decision theory literature which was presented in Section 9.< of #hapter
9. 0ariations of the model have been shown to interviewees and the version that is
presented here has been acknowledged by employees in the oil and gas industry to be
an accurate description of current practice in investment appraisal. The respondents
from #ompanies # and . said of the latest version:
HI think youBre right. CouBve basically captured it.K &#'F
and,
HI saw this diagram and I can see a lot of what IBve come across in hereK &.'
In this section, the current version of model will be presented. The model is two2
dimensional. The a*es of the model will be e*plained and 8ustified using quotes from
the research interviews. The interviewed companies will be plotted on the two2a*es.
5?9
These results can then be interpreted and this confirms many of the observations made
in section @.;. In particular it highlights the need for a study to investigate the
relationship between the use of decision analysis and organisational performance.
The *2a*is in the model relates to the number of decision analysis techniques used for
investment appraisal decisions. The quotes from the interviews presented above
clearly indicate that there is variation in the number of decision analysis tools used by
companies for investment appraisal decisions. Some are aware of all the techniques
identified in #hapter ? and use, at least partially, most of them. Some use very few.
This is well illustrated by the following respondents:
HI donBt think that I can say we use any of those techniquesK &A'F
and,
H.ecision analysis is not a common thing in this company. ItBs not a standard
that we all have to do for each decision.K &:'
The y2a*is of the model presented in this section indicates the proportion of
investment decisions that are made in each company using decision analysis
techniques. Some organisations do not use decision analysis at all even on the most
basic investment decisions to which the techniques have been applied in the literature
for many years:
H!e have no structured, scientific, way of evaluating LprospectsMK &A'.
%thers appear to use decision analysis on a limited number of decisions:
HThe drill2no drill type decision I would say we use decision analysis
techniques a lot for. I think when you are talking about real strategic
decisions, like do I make this acquisition, I think you are much further up the
qualitative end of thingsK &4'
In other companies, decision analysis is much more commonly used. As the
following e*change between the researcher and one of the interviewees from
company / indicates:
5?;
Interviewer: H%n what kind of decisions does your organisation use decision
analysis techniquesIK
Interviewee: H!ell, really, 8ust about everything we do from decisions about
drilling prospects through to development decisions and decisions about
production.K
Interviewer: H%n what kind of decisions does your organisation not use
decision analysis techniquesIK
Interviewee: HThereBs none.K &/5'
It is important to realise that on the y2a*is that the proportion does not indicate a strict
dualism. ,udgement and individual interpretation, gap filling in the absence of
complete information, and assumptions are required even when many tools are used:
HCou dip into that side and you come back and do some more numbers and
then you dip back into that side and I think that is the way it has to go because
you cannot prove mathematically that there are 5? million barrels in the
ground in a discovery. Cou have to interpret the information you have got.
And that interpretation eventually comes down to a 8udgement of somebody
which is fair and square on this side:ecause you keep asking questions and
ultimately you get down to what somebodyBs view is J somebodyBs
interpretation of a reservoir model or whatever, to which you can say no more
than that is my interpretation, that is my feeling, my view of what the thing
looks like.K &/5'F
and,
H!e like to think the thing is structured. !e like to think thereBs an ordered
trail of how we got to the decision. And we like to think, or some people like
to think, they are completely quantifiable. :ut I think it is hard to get
something that is absolutely hard quantifiable, because we are dealing with a
sub8ective process. :ecause we are donBt know all the answers J we donBt
even have all the questions.K &.'
+oreover, in all companies:
H3ltimately decisions are taken on 8udgement.K &/9'
The two a*es are measured along ordinal scales.
(lotting the interviewed companies on these two a*es, using the information obtained
in the semi2structured research interviews, then produces the model shown in figure
@.5. The pattern obtained in figure @.5 confirms the observations made in section @.;.
"irstly, it clearly supports the perception that organisations begin to use decision
5?<
analysis techniques on routine, operational decisions before introducing the
techniques corporation2wide. Secondly, it provides complementary evidence that as
companies introduce more techniques, they tend to use the techniques on more
decisions, some of which can be regarded as strategic. This implies that in companies
that use many decision analysis techniques, the decision2makers are, if not actually
using the techniques themselves, at least involved in the generation of the analysis and
in the interpretation of its results. This confirms the identification of managerial
attitude to decision analysis as a key factor in determining the use of decision analysis
techniques. Thirdly, in the model there are clearly three groups of companies &each
groups is a different colour in the figure'. This appears to confirm that organisations
are choosing not to modify which techniques they use or how they use them,
preferring instead to stay within their group. (ossible reasons for this were identified
in section @.;. These include the decision2makerBs perception of decision analysis,
which is coloured by the lack of any empirical evidence that indicates that using
decision analysis is positively associated with organisational performance.
"igure @.5 : # model of current practice
:(8 *O-*+U0IO-
:y drawing on the interview data and the insights gained from the behavioural
decision theory literature, the discussion in this chapter has established which
'roportion o"
decisions
-um#er o" decision analysis tools used
A
1
*
D,)
F,G

I,L,
.
+
M,-,O,',
A
R
0,,
5??
techniques upstream companies use for investment appraisal. This has indicated that
the gap identified by earlier research between theory and practice in decision analysis
still e*ists. This indication of current practice will be used in #hapter 7 to produce a
ranking of the companies according to the sophistication of the decision analysis tools
they use for investment decision2making. In the current chapter a model of
investment appraisal in the upstream was produced. 3sing this model reasons for the
e*istence of the gap between practice and capability were proposed. In particular, it
was suggested that managers are unconvinced about the value of decision analysis
since there is no evidence that use of the techniques leads to more successful
investment decision2making. #onsequently, organisations do not adequately resource
the introduction and use of the methods and managers regard the results as spurious
and pay limited attention to them in their investment appraisal decision2making.
Therefore, the following chapter, focuses on the third research question by using the
indication of current practice produced here to investigate the relationship between
the use of decision analysis techniques in investment appraisal decision2making and
organisational performance in the upstream oil and gas industry.
5?@
*hapter 9
,he relationship #et2een the use o" decision
analysis in investment appraisal decision-
making and #usiness success4
a non-parametric analysis
5?7
9(3 I-,RODU*,IO-
This chapter will focus on answering the third research question by e*ploring the
relationship between the use of decision analysis in investment appraisal decision2
making and business success in the upstream. %rganisational performance is
comple*, multi2dimensional and fundamental to strategic management theory and
practice &0enkatraman and $amanu8am, 56=@'. +ost researchers consider
performance the ultimate test of new concepts and theories &4eats, 56==F Schendel
and Dofer, 5676'. It is contended in this chapter that the use decision analysis
techniques and concepts in investment appraisal decision2making is a source of
competitive advantage among operating companies active in the upstream oil and gas
industry in the 34#S. This hypothesis is investigated in this chapter using non2
parametric statistical tests.
The chapter begins by establishing the type of study that will be carried out. It then
draws on the discussions of #hapters ? and @, to construct a ranking scheme. #hapter
? used the decision theory and industry literatures to present the range of decision
analysis techniques and concepts available to upstream companies for investment
appraisal and in #hapter @, it was established which of these tools and ideas
companies choose to use in investment decision2making and why. In this chapter,
using #hapter @ as the main data source, each of the upstream companies interviewed,
are ranked according to their knowledge and use of each of the techniques presented
in #hapter ?. +easures are then selected that are indicative of upstream
organisational performance and companies are ranked again this time according to the
performance criteria. Dypotheses are proposed for statistical testing. %nce the
statistical tests have been conducted, the results are analysed and discussed. The
theoretical contribution of the research to the debate between behavioural decision
theorists and decision analysts, the implications for practitioners especially to
managerial perceptions of decision analysis, the limitations of the current study and
areas for future research will be discussed in #hapter =.
5?=
9(% ,) ,C') OF 0,UDC
+ost statistical techniques can be applied in different situations that vary in the degree
of e*perimental control the researcher has and in the type of conclusion that can be
drawn &-each, 5676 p9>'. The distinction, however, is at the level of designing and
carrying out the e*periment, and not at the level of data analysis &-each, 5676 p95'.
Dence, this chapter begins by looking at the two most common types of study and
e*plores which is the most appropriate to use to investigate the relationship between
the use of decision analysis tools and concepts and organisational performance.
In the first situation, the researcher takes a random sample of companies who use
decision analysis and a second sample that do not. The performance of each is then
monitored, and the researcher notes whether those organisations that use decision
analysis score more highly than those companies that do not. In this situation, the use
of decision analysis techniques and concepts is an attribute of each company taking
part in the studyF it is inseparably attached to each of the companies. !ith such a
study, it is not possible to establish a causal relation between the use of decision
analysis tools and concepts and organisational performance but only that there is an
association between them. Thus, if a difference is found between the two samples, the
researcher is not entitled to say that using decision analysis techniques and principles
makes organisations perform better. It is quite possible in such a study that the
association could be caused by other traits or environmental factors that predispose an
organisation that uses many decision analysis techniques and concepts to perform
better. In fact, it is not even possible in such a study to rule out the possibility of good
organisational performance causing organisations to use more decision analysis tools
and concepts. Such studies are known as correlational studies.
In the second situation, the researcher controls the number of decision analysis
techniques and concepts used by organisations. The degree of usage of decision
analysis tools and ideas is a treatment rather than an attribute. The researcher takes a
random sample from the population of interest and randomly assigns each company to
one of two groups. The members of one group are given sophisticated and numerous
decision analysis techniques and concepts to use, and the others given relatively
5?6
unsophisticated and few decision analysis tools and ideas. After a certain amount of
time using their tools and concepts, the organisationsB performance is noted. Then,
even if there were, for e*ample, certain traits that cause companies to utilise many and
sophisticated decision analysis techniques and to perform well, these would balance
out by the random assignment of the companies to the two treatments. In this case, if
the users of sophisticated and numerous decision analysis techniques and principles
performed well, then this would indicate the use of decision analysis as the cause.
Thus, in the second situation, the researcher would be entitled to draw causal
conclusions, while in the first they would not. This type of study is known as an
e*perimental study &-each, 5676, pp5629>'.
In this case, an e*perimental study is impossible. This would involve finding a group
of companies in the upstream who would be willing to be assigned at random to either
using decision analysis techniques or concepts or not, on a long2term basis. )iven the
intensity of the competition between the organisations that operate in the oil industry,
one group of companies &and their shareholders' are not likely to accept that the other
group may e*perience better organisational performance for any length of timeZ
)iven this, the current research will be correlational and will aim to establish if there
is an association between organisational performance and use of decision analysis
techniques and concepts. "ollowing -each &5676 p56', the use of decision analysis
tools and concepts will be labelled the e*planatory variable and organisational
performance the response variable. -each &5676 p9>' argues that provided the
researcher acknowledges that when an e*planatory variable is handled as an attribute,
the researcher cannot conclude that any variation in the e*planatory variable
He*plainsK variation in the response variable, it is permissible for the label to be used
in correlational studies.
In the following two sections, data will be compiled and presented that indicates first,
organisationsB use of decision analysis tools and concepts and second, business
performance. In section 7.<, the statistical discussion will resume and the statistical
tests will be chosen based on the types of data that have been gathered.
5@>
9(5 RA-.I-G *OM'A-I)0 1C U0) OF D)*I0IO- A-A+C0I0 ,OO+0 A-D *O-*)',0
In #hapter ? the range of decision analysis techniques and concepts that are available
to upstream companies for investment appraisal were presented. #hapter @ indicated
which of these tools and ideas companies choose to use and why. In this section, the
two preceding chapters are used as input to construct a ranking scheme which grades
companies according to their use of decision analysis techniques and concepts, with
the higher2ranking positions being given to those companies that use a larger number
of decision analysis techniques and ideas. This ranking together with the performance
measures ranking compiled in the following section, will be statistically analysed in
section 7.?.
The techniques and concepts presented in #hapter ? comprise the toolkit currently
available to the upstream decision2maker. They vary in comple*ity from basic .#"
techniques to the more obscure option and preference theories. Some of the ideas
have been applied to the industry in the literature for many years, others only
relatively recently. !hilst for most of the tools there is software available making it
possible to automate their use, for a few there is no software package manufactured,
making manual manipulation the only option. Such factors have affected the
implementation of the techniques in companies. Dowever, #hapter @ provided
evidence of other influences, which are perhaps stronger, which have also affected
organisationsB uptake and use of decision analysis techniques. In particular, in each
company, the top managementBs attitude towards decision analysis and the corporate
culture appear to affect the e*tent to which decision analysis techniques are used.
#hapter @ confirmed the findings of earlier studies by Schuyler &5667' and "letcher
and .romgoole &566@' by providing evidence that there is a gap between practice and
capability in the e*tent to which the upstream industry use decision analysis
techniques and concepts. Dowever, it also indicated that individual companies vary in
the e*tent to which they contribute to this gap. !hilst some companies might have no
knowledge of a particular tool or concept, in others its use may well be commonplace,
and the technique or idea may be regarded as a main component of the organisationBs
investment appraisal process. "ollowing these observations, it is possible to rank
companies according to the e*tent of their usage of decision analysis tools and
philosophies. In the ranking, companies that use many decision analysis tools and
5@5
ideas will score more highly than those organisations that choose not to use decision
analysis.
The decision analysis techniques and concepts are listed below. "or ease of
presentation the tools and ideas are described roughly according to their level of
comple*ity &and, hence, ease of implementation', sophistication of output and e*tent
to which their usefulness to the industry is acknowledged in the literature. "or each
technique and concept, an indication is given of how the companies will be graded
and ranked on this criterion. !here necessary a brief outline of the tool or idea is also
provided. TechniquesNconcepts =25; used the same scoring system for ranking
companies. This is e*plained in the discussion of tool 5;.
3 Auantitative analysis( This is used here to refer to the calculation by analysts of
decision2making criteria such as payback, rate of return &:uckley, 9>>>' or
discounted profit to investment ratio &Digson, 566?'. The calculation of these
criteria are recognised by many analysts to be the most basic type of investment
appraisal companies can undertake since the measures are simple to calculate,
include no e*plicit recognition of the e*istence of risk and uncertainty and hence,
their output is primitive &for e*ample, /ewendorp, 566@'. Two points will be
assigned to companies that calculate these criteria routinely in their investment
appraisal process. %ne point will be given for partial implementation, and 1ero
for non2usage.
% olistic vie2. #hapter ? indicated that for companies to make _properB decisions
it is essential that they adopt a holistic view of the total cumulative net effect of
the consequences of the decision currently under consideration. "or e*ample, for
any upstream investment decision, there must be an estimate of the timing and
cost of the abandonment of the facilities and the cost and timing implications of
any environmental protection measures that may need to be taken. "or a full
discussion refer to :all and Savage &5666'. The need for upstream organisations
to adopt a holistic perspective is well documented &Simpson et al., 9>>>F
/ewendorp, 566@' and simple to achieve. Those companies that adopt a holistic
view of the total cumulative net effect of the consequences of the decision being
taken will be assigned two points. The companies that recognise the necessity to
5@9
do so but mostly do not will be given one point. /o points will be given to
companies that do not recognise the need to take a holistic perspective.
5 Discounted cash "lo2 techni@ues. As discussed in #hapter ?, the timing
characteristics of upstream pro8ects are such that there is an historical average, in
the /orth Sea, of about seven years between initial e*ploration e*penditure and
commitment to develop, with another three or four years to first production and
then twenty years of production revenues before abandonment e*penditure.
$ecognition of this, and of the time value of money, means that .#" techniques
&see, for e*ample, :realey and +yers, 566@' must be used by upstream
companies. .#" is relatively easy to conduct, its usefulness to the upstream well
documented and the output it produces simplistic. Two points will be assigned
where companies use .#" techniques routinely in their investment appraisal
process and have appropriate training for employees in how to use the tool. %ne
point will be given for partial implementation, and 1ero for non2usage.
7 Risk and uncertainty. In Section 9.9 of #hapter 9 the literature review indicated
that there are numerous definitions of risk and uncertainty presented in the
literature and that the conceptualisation that decision2makers adopt affects the
method of coping that they &and their organisation' adopts. #learly, then
companies ought to have corporate definitions or, at least, a tacit organisational
understanding of the terms risk and uncertainty, which are complementary to their
approach to investment appraisal. $isk and uncertainty have received much
attention in the industry literature and numerous definitions proposed for
organisations to select from. The definitions ought to be easily applied via
training or workshops.
#ompanies will be assigned two points if they have organisation2wide definitions
or understandings, of the terms that fit with their approach to investment appraisal.
%ne point will be given if they have any definitions or tacit understanding at all
and no points will be allocated if the company has no definition or understanding
of the concepts of risk and uncertainty.
5@;
8 Monte *arlo "or prospect reserves( #hapter ? provided a discussion of the
benefits of using risk analysis via +onte #arlo simulation to generate a
probabilistic estimate of recoverable reserves. Simulation has been applied to
reserve evaluation in the literature for many years and software now e*ists to
make this process relatively simple. The output produced by the simulation is a
probability distribution of the recoverable reserves. %rganisations that adopt this
approach for prediction of recoverable reserves are e*plicitly recognising the
e*istence of risk and uncertainty in these estimates. #ompanies will be given two
points if they routinely use +onte #arlo simulation to generate estimates of
prospect reserves. %ne point will be assigned to those organisations that
occasionally used the technique and no points will be allocated for non2usage.
: p3&, p8& and p6& reserve cases "or economic modelling( Three reserve cases
should be used as input into their economic modelling since they are
representative of the best, worst and most likely outcomes.
Those companies that use the three reserve cases specified above will be assigned
two points. !here organisations occasionally use the three cases but usually only
use one reserves case for economic modelling, the company will be given one
point. !hen the economic models are purely constructed on one reserves case,
these companies will be given no points.
9 )M? via decision tree analysis( The value of calculating an A+0 through a
decision tree is widely acknowledged in both the industry and decision theory
literatures. Two points will be assigned to companies that use decision tree
analysis to calculate an A+0 routinely in their investment appraisal process and
have appropriate training for employees in how to construct decision trees and
calculate A+0s. %ne point will be given for partial implementation, and 1ero for
non2usage.
; 'ro#a#ilistic prospect economics( Since, firstly, the technology e*ists for
automated probabilistic economics, and secondly, it is widely documented that
economic variables are volatile, companies ought to be producing probabilistic
prospect economics. Dowever, producing probabilistic prospect economics can be
5@<
comple* since there are many economic variables to consider with many
dependencies between them. The necessity to produce probabilistic prospect
economics is receiving increasing attention in the industry literature &for e*ample,
:ailey et al., in pressF Simpson et al, 5666F -amb et al., 5666F Snow et al., 566@'
and the output that is produced e*plicitly recognises the e*istence of risk and
uncertainty in economic estimates.
6 'ro#a#ilistic production reserves( "ollowing the same rationale as =, companies
ought to be e*plicitly recognising the risk and uncertainty in estimating
production reserves by using probabilistic analysis.
3& 'ro#a#ilistic production economics( "ollows from = and 6.
33 'ort"olio theory. +arkowit1

has shown how a diversified portfolio of uncertain
opportunities is preferable to an equal investment in a single opportunity, or
restricted portfolio of opportunities, even if the diversified portfolio contains
pro8ects that are more risky than any other pro8ect in the restricted portfolio.
Authors such as :all and Savage

&5666' have taken this concept and applied it to
the upstream oil and gas industry, showing how diversification in terms of
geographic or geological setting, in product pricing mechanism &gas versus oil',
production profile timing, political environment, and the avoidance of specific
niches can, when the alternatives have a negative correlation, have a positive
impact on the riskNreward balance of the companyBs investments. It is clear, then,
that when considering an incremental investment, its impact on the total portfolio
should form an important factor in the decision2making.
(ortfolio theory has been applied to the upstream industry in the literature for
several years and whilst the concepts are relatively simple the mechanics of the
technique are comple*. This makes implementation more difficult. There is
software produced to automate its use.
3% Option theory. There are four significant characteristics of most of the decisions
taken in the upstream. These are: they form part of a multi2stage decision process
&figure ?.5'F they are, to a large e*tent, irreversibleF there is uncertainty associated
5@?
with most of the input parameters to the decision analysisF and a decision2maker
can postpone the decision to allow the collection of additional data to reduce risk
and uncertainty. These characteristics mean that traditional .#" techniques can
be modified through the application of option theory &see, for e*ample, .i*it and
(indyck, 566= and 566<' to assign credit to an opportunity for being able to assess
and to avoid the downside uncertainty involved in a decision by aborting or
postponing that decision until certain conditions are met. +any companies having
been doing this to a limited e*tent, perhaps without realising that it represented the
application of option theory, through the use of decision trees. The simple
representation given by the decision tree in figure 7.5, illustrates the benefit of
minimising e*penditures by realising that a discovery may be too small to be
economic and e*ercising the option of limiting investment to e*ploration and
appraisal seismic and drilling, and waiting until commercial considerations &price,
costs, ta*es' change and the field becomes economic, or not developing at all,
rather than developing at a loss. .i*it and (indyck

&566<' outline more rigorous
mathematical techniques for assessing the option value of the uncertainty in an
investment over which one has the ability to delay commitment, but the principle
is the same.
The value of applying option theory to the oil industry has still to be proven. It
has only recently begun to attract significant attention within the industry
literature and there is no software currently available to automate its use.
"igure 7.5 : # decision tree
#ostNvalue
**
**


#ostNvalue

*hance
#ostNvalue
K
JM
JM

Dry
ole
Discovery
-
o
n
-
c
o
m
m
e
r
c
i
a
l
*ommercial
** V ** T **
** V ** T **
** V ** T **
** V ** T **
** V ** T **

Opportunity )M? M ==
(6>
(?>
(5>
5@@
35 're"erence, or utility, theory( As indicated in #hapter ?, this aspect of decision2
making recognises the fact that companies &or, indeed, decision2makers within
companies' do not all have the same attitude towards money. "or e*ample, a
smaller company will be much less able to sustain losses than a larger company,
and will therefore be much more wary of risky pro8ects with downside risks that
could bankrupt the company.
!hilst preference theory has been widely applied to the industry in the literature,
its value is questionable. There are difficulties in obtaining preference curves and
in the construction of corporate preference curves. There is, however, some
software available to automate the technique.
"or techniques =25;, two points will be assigned where the technique is used
routinely in organisations for investment appraisal and appropriate training is
given to staff. %ne point will be allocated for partial implementation and 1ero
points for non2usage.
37 Aualitative and @uantitative input( #hapter @ showed that few, if any, decisions
are based solely on quantitative analysis. It seems that decision2makers usually
ultimately lack the faith to act purely on the basis of the quantitative output. In
such cases qualitative influences, such as habit, instinct, intuition or imitation of
others, are also used. !hichever, the reasons for disregarding quantitative
analysis, or amending it, are to do with 8udgements of 8udgements, where the
original 8udgements are inputs to quantitative analysis and the 8udgements of
8udgements pertain to strength of belief. #ompanies ought to have a systematic
method for documenting and critically e*amining or calibrating their qualitative
input. This would ensure transparent decision2making in alliances and
partnerships.
+uch of the decision theory and industry literature fails to acknowledge the need
for organisations to manage the qualitative and quantitative interface. There is no
commercial software available that allows qualitative factors first, to be
transparently included in the generation of the quantitative analysis, second, to be
e*plored and modelled. Through recently established collaborative relationships,
5@7
the ma8or players &#SI$%, +erak, )affney2#line ^ Associates, !ood +acken1ie
and ./0' are now working together in an attempt to deliver to the upstream
decision2maker such software.
In the company ranking, organisations that manage the qualitativeNquantitative
interface e*plicitly perhaps by using soft methods &for e*ample, through simple
multi attribute rating techniques' will be assigned two points. %ne point will be
allocated to organisations that acknowledge the e*istence of both quantitative and
qualitative inputs to the decision2making process but do not try to manage their
interaction overtly. /o points will be given to companies that do not recognise
nor manage the qualitative and quantitative input to the decision2making process.
This list represents the collection of decision analysis tools and concepts presented in
#hapters ? and @. The techniques and ideas do not represent alternatives but rather a
collection of analysis tools and principles to be used together to encourage more
informed decision2making. The organisation that utilises the full spectrum of these
techniques and ideas is perceived to be adopting a transparent approach to decision2
making that manages the full range of qualitative and quantitative aspects of the
process.
*OM'A-C
*RI,)RIA A 1 * D ) F G I L . + M - O ' A R 0 ,
Guantitative analysis 3 % % % % % % % % % % % % % % % % % % %
Dolistic view 3 3 % % % % % % % % % % % % % % % % % %
.iscounted cash flow 3 % % % % % % % % % % % % % % % % % % %
$isk and uncertainty
definitions
& & & & 3 & & 3 & % 3 & % & % 3 % 3 3 3
3se +onte #arlo for
prospect reserves
& & 3 % % % % % % % % 3 % % % % 3 % % %
Take p5>,p?>,p6> reserve
cases
& & 3 & & & 3 & 3 & & 3 3 % % 3 3 % % %
#alculate an A+0, via
decision tree analysis
& & 3 % 3 % & % % % % % % % % % 3 % % %
3se +onte #arlo for
prospect economics
& & & & & & & & & & & & & 3 & 3 & & 3 &
3se +onte #arlo for
production reserves
& & & & & & & & & & & & & & & & & 3 & &
3se +onte #arlo for
production economics
& & & & & & & & & & & & & & & & & & & &
(ortfolio theory & & & & & & 3 & 3 & 3 % & 3 3 % 3 3
%ption theory & & & & & & & & & & & 3 & 3 & 3 3 & 3 3
(reference theory & & & & & & & & & & & & 3 3 & & 3 & & 3
Gualitative and quantitative & & & & & & & & & & & & & & & & & & 3 3
,O,A+ 5 8 6 3& 3& 3& 3& 33 3% 3% 3% 35 38 38 38 38 38 3: 39 39
Table 7.5: <ankin! of companies 'y use of decision analysis techni"ues and concepts (redE5 points,
!reenE1 point and 'lueE9 points
3sing #hapter @ as the main data source, the companies that were interviewed were
ranked according to the criteria specified above. The result of this ranking is shown
5@=
in table 7.5. The red squares are used to indicate where companies were assigned two
pointsF the green squares one point and blue no points. "or each of the techniques and
concepts, where there were numerical ties according to the criteria detailed above, the
tie was broken on the basis of other material from the interviews, which was not
available for every company &and therefore, not included as an overall rank measure'.
"or e*ample, the tie between companies S and T was broken on the basis that
company T applied decision analysis software company2wide whereas in organisation
S access to such software was restricted. The gap between practice and capability
identified in #hapter @ is shown e*plicitly in the table.
The following section proposes the criteria that will be used to measure organisational
performance. These measures together with table 7.5 will be used in section 7.? for
the statistical analysis of the association between organisational performance and use
of decision analysis techniques and concepts.
9(7 RA-.I-G *OM'A-I)0 1C ORGA-I0A,IO-A+ ')RFORMA-*)
In this section, financial measures will be selected that are indicative of organisational
performance in the upstream. The upstream shares with other industries such as the
pharmaceutical and aeronautics industries specific characteristics that make assessing
performance particularly challenging. Dence, financial criteria that are not typically
associated with organisational performance are more pertinent in this case. There are
also other unique measures, which indicate success in the oil industry. These will be
included in the assessment of organisational performance in the upstream.
(apadakis &566=' comments that despite the fact that performance is the most critical
and frequently employed variable in strategy research &for e*ample, Dambrick and
Snow, 5677', its theoretical aspects have not been adequately developed and tested
&4eats, 56=='. #ompounding this, measuring organisational performance in different
industries, and even in different samples, presents distinct challenges. #onsequently,
previous researchers studying the decision2making process have used various and
different criteria to assess organisational performance &0enkatraman and $amanu8am,
56=7F .ess and $obinson, 56=<'. "ollowing this trend, the current study uses a
variety of measures to assess organisational performance. The choice of these criteria
5@6
is limited by two factorsF firstly, by the data that is available. Some oil companies
appear to report e*tensively whereas others only publish that which they are required
to do by law. "urthermore, despite the trend toward using non2financial measures
&such as customer acquisition, retention and satisfaction, employee satisfaction and
organisational learning &#hang and +organ, 9>>>F van de 0liet, 5667F Dalley and
)uilhorn, 5667F +anagement Accounting, 5667F -othian, 56=7F Darper, 56=<' to
measure company performance, such criteria are either inappropriate for the upstream
companies under investigation since several are integrated oil companies with both
upstream and downstream business and hence issues of customer acquisition are
irrelevant, or are not widely reported by the oil companies. Secondly, the selection of
measures is restricted because investment decision2making in the oil industry is
unique. $ecall, from #hapter ? that the oil industryBs investment decisions are
characterised by a long payback period. In the case of e*ploration and development
decisions, this time2period can be up to fifteen years. Therefore, to some e*tent,
companiesB performances now are dependent on decisions taken many years ago when
the industry did not routinely use decision analysis &Section @.; of #hapter @'. So to
investigate the relationship between the use of decision analysis and oil companiesB
business success, it is important that measures are selected that reflect the effect of
recent decision2making. In the oil industry, this is best acknowledged by measures
that indicate the successfulness of recent e*ploration decisions. This includes, for
e*ample, !ood +acken1ieBs estimate of a companyBs total base value which is
calculated by the values of commercial reserves, technical reserves &as defined by
!ood +acken1ie' and the value of currently held e*ploration and !ood +acken1ieBs
assessment of its potential.
Therefore, the following criteria will be used in this study to be indicative of
organisational performance in the upstream. Aach measure is reviewed below with
particular attention being focussed on the conclusions that the researcher will be able
to draw by using the criterion.
,he volume o" #ooked reserves or proved reserves !'R$( (roved reserves are
reserves that can be estimated with a reasonable certainty to be recoverable under
current economic conditions. #urrent economic conditions include prices and
57>
costs prevailing at the time of the estimate. (roved reserves must have facilities to
process and transport those reserves to market, which are operational at the time
of the estimate or there is a reasonable e*pectation or commitment to install such
facilities in the future. In general, reserves are considered proved if the
commercial producibility of the reservoir is supported by actual production or
formation tests. In this conte*t, the term proved reserves refers to the actual
quantities of proved reserves and not 8ust the productivity of the well or reservoir
&Society of (etroleum Angineers et al., 9>>>'. "or the company performance
ranking, the volume of proved reserves will be used as a pro*y for the si1e of the
organisation and as an indicator of recent, past results in investment decision2
making.
/ood MackenEieNs estimate o" each companyNs total #ase value !,1?$( As
indicated above, !ood +acken1ie calculate this measure by summing the values
of a companiesB commercial reserves, technical reserves and the value of currently
held e*ploration and an assessment of its potential. "or the organisational
performance ranking, this measure is particularly attractive as it e*plicitly
includes an assessment of the success of recent, past investment decision2making.
Dowever, !ood +acken1ie only publish an estimate of each companyBs 3.4.
T:0 complied from 34#S data. This is an obvious weakness as some companies
choose not to operate in mature basins like the 34#S or are scaling down their
operations due to the high costs involved in operating in the 3.4. &Section ;.; of
#hapter ;'. #urrently, however, no other group of analysts produces a similar
measure reflecting worldwide T:0 &or an equivalent criterion that reflects the
value of recent e*ploration'. Acknowledging then the weakness of the measure,
but recognising there is no alternative criterion, this research will use the 3.4.
T:0 produced by !ood +acken1ie in combination with other criteria that are
indicative of worldwide performance.
Return on e@uity !RO)$. $%A is defined as the equity earnings as a proportion
of the book value of equity. It is a measure of overall performance from a
stockholderBs perspective and includes the management of operations, use of
assets and management of debt and equity. $%A measures the overall efficiency
of the firm in managing its total investments in assets. In the conte*t of the
upstream, this measure does not include the effects of decisions taken in the recent
575
past. &In fact, the opposite since although the measure acknowledges the
monetary investment of recent decisions, the long payback period means that
returns have not yet been earned'. The measure is included in the performance
ranking for comparison with the criteria that do reflect the effects of recent
decision2making and as an indicator of the results of past investment decision2
making.
Market capitalisation !M*$( +# is defined to be the total value of all
outstanding shares in sterling. It is used in the performance ranking as a measure
of corporation si1e.
-um#er o" employees !-O)$( The /%A is used in the performance ranking as a
relatively coarse indicator of both past success and anticipated future success in
selecting and gaining access to the best investment opportunities.
'rice earnings !')$ ratio. The (A ratio relates the market value of a share to the
earning per share and is calculated by:
(rice earnings T +arket value per share
Aarnings per share
The ratio is a measure of market confidence concerning the future of a company.
In particular, it is used in the performance ranking as an indicator of growth
potential, earnings stability and management capabilities. The higher the price
earnings ratio, the greater the market believes is the future earning power of the
company. This measure does not e*plicitly include the effects of decisions taken
in the recent past but it is used here for comparison with the criteria that do.
'rudential 0ecurities ranking !'0R$( In 9>>>, (rudential Securities carried out
an energy industry benchmarking study that used nine variables to rank the ma8or
oil companies. The variables which they considered were: production incomes,
quality of earnings, cash flow, production and replacement ratios &e*cluding
abandonment and disposal', finding and development costs &e*cluding
abandonment and disposal', discounted future net cash flows, upstream returns,
ad8usted production costs and depreciation, depletion and amorti1ation e*penses.
Some of the measures above, such as proved reserves, are influenced by the si1e
of the organisation, since (S$ is based on financial measures, small and large
579
companies can be compared and hence it provides a useful indication of business
success which independent of organisational si1e.
!here possible the data used to calculate each measure will be based on the latest
figures released by companies. In the case of the 3.4. T:0 criterion, the data used
will be based on !ood +acken1ieBs latest estimates produced in April 9>>>. "or the
$%A, 566= figures will be used, as these are the most recent complete data set
available. (revious strategy research &for e*ample, )oll and $asheed, 5667F )rinyer
et al., 56==F (apadakis, 566=' averaged performance criteria over a five year period,
to decrease the chance of a one2year aberration distorting the results. !hilst in
general this is good research practice, in this case this is not appropriate since this
would involve aggregating criteria across time periods where decision analysis was
not used routinely by the ma8ority of the participants &Section @.9 of #hapter @'.
All the measures described above, with the e*ception of the 3.4. T:0, are indicative
of each companyBs worldwide performance yet, typically, the respondents were
employees working within 3.4. offices. Dowever, the researcher does not perceive
this to present a problem since each interviewee was specifically asked to comment
on the techniques that they were aware that their organisation used to evaluate
investment opportunities worldwide and how they perceived these tools and the
overall process to work organisation2wide. Therefore, the researcher is confident that
the observations from the interviewees are not significantly biased by their place of
work and that it is acceptable to rank the companies using measures indicative of
worldwide performance.
+ost of the companies included in the analysis have both up and downstream
operations. Since very few of the companies differentiate between the two in their
publication of financial data some of the measures chosen &for e*ample, +#, (A and
/%A' reflect organisational performance in both areas. Since, arguably the
downstream business is dependent on successful decision2making in the upstream,
this is only of slight concern. Dowever, the criteria that reflect only upstream
performance &($, T:0 and (S$' will be given more attention.
57;
The companies that were interviewed were ranked according to the performance
criteria selected above. &The data gathered to construct this ranking came from a
variety of web sites: (rudential Securities &http:NNwww.prudentialsecurities.com',
world vest base &http:NNwww.wvb.com', wrights investors service
&http:NNwww.wsi.comNinde*.htm', financial times &http:NNwww.ft.com', hooverBs on
line business network &http:NNwww.hoovers.com' and datastream,
http:NNwww.datastreaminsite.com.'. The results of this ranking are presented in table
7.9. &#ompanies are listed worst2best performers, top2bottom for each criterion'.
.ata for some categories and companies is incomplete because the information
proved impossible to locate. The ranking is used in the following sections as an
indication of organisational performance.
,1? M* 'R -O) RO) ') '0R
#
.
(
)
:
4
I
"
T
/
S
%
-
$
.
(
A
A
/
:
I
S
,
T
+
$
%
A
.
,
A
/
4
:
I
S
T
(
-
$
%
#
(
.
A
A
/
S
I
)
,
T
:
4
+
%
$
-
.
(
A
/
4
A
)
-
S
I
#
T
:
+
,
$
%
.
/
,
I
/
:
A
T
+
A
%
4
$
I
:
/
%
T
+
-
$
S
Table 7.9: <ankin! of companies 'y performance criteria (Companies are listed +orst-'est
performers, top-'ottom for all criteria
9(8 'RO'O0I-G ,) C'O,)0)0 A-D 0)+)*,I-G ,) 0,A,I0,I*A+ ,)0,0
57<
This section uses the discussion in section 7.9, the company ranking constructed in
section 7.;, and the organisational performance ranking compiled in the preceding
section, to ascertain which statistical tests are the most applicable to use to investigate
the relationship between the use of decision analysis and organisational performance.
The appropriate null and alternative hypotheses are proposed for empirical testing.
The choice of statistical test is always comple* and governed, primarily, by the type
of data available and the question being asked &-each, 5676 p95'. $esearchers must
assess whether their data are ordinal or categorical, independent or related and pertain
to one sample or several samples. :y e*ploring these issues, statisticians such as
-each &5676 p99' argue, researchers ought to be able to at least reduce their choice of
statistical test.
In the current study, whilst for the ranking of companies by their use of decision
analysis techniques and concepts, categorical and ordinal data are available, when
these data are e*pressed categorically there are many ties in the data. "or some of the
performance measures, it is only possible to access ordinal data.
The data in each ranking are independent. This claim can be substantiated in two
ways. "irstly, in the oil industry, the performance of one company is not significantly
influenced by the success of another. All companies are sub8ect to the fluctuations of
the oil price and to the vagaries of depositional environment. Secondly, in #hapter @
it was shown that companies do not significantly influence each other to adopt new
techniques or concepts. The investment appraisal approach that is adopted in the
company is more likely to be affected by internal organisational factors such as
managementBs perception of decision analysis and the corporate culture than the
techniques or approach used by other companies.
In this case, the e*planatory variable, the organisationsB use of decision analysis
techniques and concepts, has multiple levels and hence, the problem should be
regarded as a Hseveral sampleK problem.
This process highlights three tests as being applicable in this case: 4endallBs test for
correlation, SpearmanBs test for correlation and the 4ruskal !allis test. "irst, consider
57?
the two correlation tests. Since the two tests rarely produce different results &-each,
5676 p569' and the researcher is familiar with the Spearman correlation test, it will be
used here. The procedure for carrying out the Spearman test for correlation is
outlined in Appendi* ;. The null and alternative hypotheses that will be tested using
the Spearman test for correlation for each performance measure are:
D5
>
: There is no or a negative relationship between the ranking of sampled companies
with respect to the performance measure under investigation and the ranking of
the sampled companies with respect to decision analysis sophistication in
investment appraisal.
D5
5
: There is a positive relationship between the ranking of sampled companies with
respect to the performance measure under investigation and the ranking of the
sampled companies with respect to decision analysis sophistication in investment
appraisal.
The 4ruskal !allis test is a direct generalisation of the !ilco*on $ank Sum test to
three or more independent samples. The test attempts to decide whether the samples
of scores come from the same population or from several populations that differ in
location. It assumes that the data are independent and ordinal. The procedure for
carrying out the test is outlined in Appendi* <. Since ($ and T:0 are two of the
criteria which are most indicative of the results of recent, past investment decision2
making &section 7.<', 4ruskal !allis tests will only be carried out on them &there is
insufficient data for a 4ruskal !allis test for (S$'. The null and alternative
hypotheses to be tested will be:
D9
>
: The T:0 &or ($' of each company is independent of the decision analysis
sophistication rank achieved by each company
D9
5
: The T:0 &or ($' of each company come from populations that differ in location
according to the rank achieved by each company in the assessment of decision
analysis sophistication.
If a significant result is achieved with this test for either or both of the criteria the
locus of the difference will be identified by carrying out multiple comparisons using
the !ilco*on $ank Sum test. This test is also outlined in Appendi* <.
57@
The following section investigates these hypotheses by calculating the appropriate test
statistics.
9(: R)0U+,0
In this section, the results of the statistical tests are presented and the null hypotheses
are accepted or re8ected as appropriate.
The Spearman tests for correlation were carried out and the results are presented in
table 7.;. Inspection of this data indicates that < of the 7 criteria provide statistically
significant relationships, at a level of ?E or less. Dighly significant positive
correlations are produced between the performance criteria T:0 and ($, and use of
decision analysis tools and ideas. There is also a strong significant positive
correlation between +# and (S$, and the use of decision analysis techniques and
concepts. There are only weak positive correlations between the categorisation of
decision analysis and the rankings of $%A and (A and neither is significant at any
level. Therefore, the null hypotheses &D5
>
' for +#, T:0, ($ and (S$ can be re8ected
and the alternative hypotheses &D5
5
' accepted. "or (A and $%A, it is not possible to
re8ect the null hypotheses &D5
>
'.
?ARIA1+)
0')ARMA-
*ORR)+A,IO-
*O)FFI*I)-,
+)?)+ OF
0IG-IFI*A-*)
'R $T>.7>5, nT5< (`>.>>?
M* $T>.?;=, nT5; (`>.>?
,1? $T>.@??, nT5@ (`>.>>?
-O) $T>.;=9;, nT57 (`>.5
RO) $T>.9?9, nT57 /NA
') $T>.96@, nT5; /NA
'0R $T>.@, nT6 (`>.>?
Table 7.; : Spearman correlation coefficients 'et+een performance ,aria'les and use of decision
analysis
"or the 4ruskal !allis test for ($ the test statistic 4 is calculated to be =.5<9=. There
are 9 degrees of freedom and hence this is significant at the ?E level. The null
hypothesis &D9
>
' for ($ can then be re8ected and, by implication, the alternative
hypothesis &D9
5
', that there are differences between the samples, accepted. To
577
determine the locus of this difference, multiple comparisons are made using the
!ilco*on $ank Sum test, with the null hypothesis each time being that the samples
were from the same population, and the alternative hypothesis being that the samples
were from several populations that differ in location. The !ilco*on $ank Sum test
indicates that those companies that are ranked in the top ten in the sophistication of
decision analysis ranking all have similar ($s. Dowever, their ($s are significantly
bigger than those companies that were placed between 55 and 5< in the decision
analysis sophistication ranking. &All calculations are shown in Appendi* <'.
#arrying out the 4ruskal !allis test for T:0 in e*actly the same way produces
similar results. The test statistic 4 is equal to 7.;7. There are 9 degrees of freedom
and therefore this is significant at the ?E level. The null hypothesis &D9
>
' can then be
re8ected and, by implication, the alternative hypothesis &D9
5
', that there are
differences between the samples, accepted. To determine the locus of this difference,
multiple comparisons are made using the !ilco*on $ank Sum test, with the null
hypothesis each time being that the samples were from the same population, and the
alternative hypothesis being that the samples were from several populations that differ
in location. The !ilco*on $ank Sum test indicates that those companies that
achieved a mid2low decision analysis ranking position &i.e. between @
th
and 5@
th
' do
not have different T:0s from each other. Dowever, their T:0s are lower than those
companies that achieved higher positions in the decision analysis ranking &in the top
?'.
The ma8ority of the empirical results then, suggest that there is a positive link between
the use of decision analysis and organisational performance. The lack of a
statistically significant positive correlation between the use of decision analysis and
$%A and (A ratio is not une*pected. As indicated in section 7.<, these two measures
are both indicative of historical decision2making and hence, decision2making when
decision analysis was not routinely used by many upstream companies &Section @.; of
#hapter @'. Dowever, the Spearman correlation coefficients for T:0 and ($, the two
criteria that are most indicative of upstream performance and that also take into
account recent decision2making and, hence, decision2making using decision analysis,
were significant at the >.?E level. In the case of ($ and T:0, the 4ruskal !allis and
!ilco*on $ank Sum tests confirmed that those companies that use sophisticated
57=
decision analysis methods were more likely to have high T:0s and high ($s. The
statistically significant positive correlation between (S$ and use of decision analysis
indicates that the relationship is independent of the si1e of the company. Dence, the
researcher is confident in asserting that there is an association between the use of
sophisticated decision analysis techniques and organisational performance. The
following section discusses these results within the conte*t of the decision theory and
organisational performance literature.
9(9 DI0*U00IO-
"rom these results, it is evident that there is an association between successful
companies and the use of sophisticated decision analysis techniques and concepts.
Dowever, before discussing this association further, it is important to acknowledge
that the nature of the study prevents the researcher from concluding that use of
decision analysis alone improves organisational performance. Indeed some writers
such as !ensley &5666 and 5667' would argue that business success cannot under any
circumstances, be ascribed to any one variable since its determinants are too comple*
for such a simple e*planation. +oreover, it will be impossible to ascertain, whether
using decision analysis tools precipitates business success or if it is once success is
achieved that organisations begin to use decision analysis techniques and concepts.
Dowever, despite these limitations, it is possible to draw the following conclusions
from the current study. "irstly, it is clear that the decision process matters and
secondly, and fundamentally, that decision analysis can be e*tremely valuable to the
upstream oil and gas industry in investment appraisal decision2making and, arguably
therefore, to other industries with similar investment decisions. +anagers have the
power to influence the success of decisions, and consequently the fortunes of their
organisations, through the processes they use to make crucial decisions. :y drawing
on the literature review of #hapter 9, the following paragraphs conte*tualise the
results of the statistical tests.
Section 9.? of #hapter 9 identified two areas of empirical literature on the relationship
between the decision2making process and effectiveness. The first demonstrated
relationships between features and types of strategic planning and firm performance.
In particular the research to date has tended to focus on the effects of
576
comprehensivessNrationality and formalisation of the decision2making process on the
performance of the company. #hapter 9 also established that use of decision analysis
implies comprehensivenessNrationality and formalisation of the decision2making
process. Dence, the results of the previous section appear to corroborate the stream of
research that suggests that either high levels of performance produce enough
resources to help organisations make more rational decisions, or that more rational
decisions may lead to better performance &,ones et al., 566;F Smith et al., 56==F .ess
and %riger, 56=7F )rinyer and /orburn, 567727='. :y implication, then, the findings
seem to refute the research that suggested that superior performance may lower the
e*tent to which organisations engage in rationalNcomprehensive, formalised decision2
making &:ourgeois, 56=5F #yert and +arch, 56@;F +arch and Simon, 56?='.
The second area of empirical research identified in Section 9.? of #hapter 9 related to
the impact of consensus on organisational performance. It was argued in that chapter
that use of decision analysis encouraged communication and helped to build
consensus amongst organisational members. As such the findings of section 7.?
appear to confirm the research of :ourgeois &56=5' and .ess &56=7' and others
&Dambrick and Snow, 56=9F #hild, 567<' who suggested that either business success
leads to higher levels of consensus, or that high levels of consensus encourage better
organisational performance. Simultaneously, the results seem to dispute those of
)rinyer and /orburn &567727=' and others &Schweiger et al., 56=@F .e !oot et. al.,
567727=' who found evidence of a negative correlation between consensus and
performance, and those of !ooldridge and "loyd &566>' who found no statistically
significant relationship at all.
9(; *O-*+U0IO-
In conclusion, the results from the current study provide some insight into the
association between performance and the use of decision analysis in investment
appraisal. The analysis presented above shows strong positive correlations between
the use and sophistication of decision analysis techniques and concepts used and
various measures of business success in the upstream. This is consistent with the
proposition that sophistication in the use of decision analysis in investment appraisal
decision2making is a source of competitive advantage in organisations that operate in
5=>
the oil and gas industry. The theoretical contribution of this research to the debate
between behavioural decision theorists and decision analysts, the implications for
practitioners especially to managerial perceptions of decision analysis, the limitations
of the current study and areas for future research will be discussed in the following
chapter.
5=5
*hapter ;
*onclusion4 #et2een <e=tinction #y instinct>
and <paralysis #y analysis>
5=9
;(3 I-,RODU*,IO-
This chapter will draw on those that precede it to answer the research questions posed
in the first chapter of this thesis. It will demonstrate how, through the utilisation of
qualitative methods and statistical analysis, the current study has produced research
that has made a valuable contribution firstly, to the decision theory and oil and gas
industry literatures and secondly, to oil industry practitioners.
:y drawing on the insights gained through the conduct of the literature review,
research interviews and data analysis stages of the current study, the chapter begins by
answering each of the three research questions in turn. The main conclusions of the
research are then conte*tualised in the decision theory and oil industry literatures.
This highlights how the research has contributed to one of the current debates in these
literatures by providing evidence of a link between the use of decision analysis in
investment appraisal decision2making and good organisational performance.
Implications of the study and recommendations to practitioners follow. The chapter
concludes by identifying directions for future research.
;(% ,) R)0)AR* AU)0,IO-0 R)?I0I,)D
In #hapter 5 the three research questions that the current study aimed to e*amine were
proposed. "ocussing on each question in turn in this section these questions and their
motivation will be e*amined. Attention will be concentrated on how, through the
utlisation of the combination of qualitative methods and statistical analysis, the
research presented in this thesis can be used to answer these questions.
The first research question aimed to establish which of the decision analysis
techniques presented in the decision theory and industry literatures, are the most
appropriate for upstream oil and gas companies to utilise in their investment decision2
making. This question was motivated by the recognition that there are many decision
analysis techniques and concepts presented in the decision theory and industry
literatures. Some of these have been applied to the upstream in the industry literature
since the early 56@>s. A few have only recently began to attract attention. Still others
5=;
have yet to be considered in the conte*t of the upstream. Dowever, previously in the
decision theory and industry literatures, authors had tended to describe the application
of a single technique to either a real or hypothetical decision situation &Dammond,
56@7F Swalm, 56@@'. !hilst such accounts provide useful insights, they also implied
that using particular decision analysis technique in isolation, would provide the
decision2maker with best possible perception of the risk and uncertainty associated
with a decision. Cet, as indicated in #hapter ?, in reality, each tool has limitations
&-efley and +organ, 5666', some that are inherent, others which are caused by a lack
of information or specification in the literature. As such, the knowledge that the
decision2maker can gain from the output of one tool is limited. Therefore, a
combination of decision analysis techniques and concepts should be used. This would
allow the decision2maker to gain greater insights and, hence, encourage more
informed decision2making. Some writers had recognised this and presented the
collection of decision analysis tools that they believed constituted those that decision2
makers ought to use for investment decision2making in the oil and gas industry &for
e*ample, /ewendorp, 566@'. Dowever, as indicated above, new techniques, such as
option theory, have only recently been applied to the industry and clearly, these
previously presented approaches required modification.
The research presented in this thesis addressed this first question by drawing on the
decision theory and industry literatures to ascertain which decision analysis tools are
the most appropriate for upstream oil companies to use for investment appraisal
decision2making. This involved firstly, identifying the whole range of techniques that
are available and, secondly, deciding which of these are the most appropriate for
upstream investment decision2making. This meant careful consideration of factors
such as the business environment of the upstream and the level and type of
information used for investment decision2making in the industry.
Through this process, the research identified the following decision analysis
techniques as particularly useful for upstream investment decision2making: the
concepts of e*pected monetary value and decision tree analysis, preference theory,
risk analysis, portfolio theory and option theory. Then by drawing again on the
decision theory and industry literatures, and also on insights gained at conferences
and seminars, a 62step investment decision2making process was presented. This
5=<
provided an illustration of how these tools can be used together in the particular
decision situation where an upstream company is considering whether to drill an
e*ploration well in a virgin basin at an estimated cost of S5> million. The approach
was summarised in figure ?.59 and this is reprinted here.
5. Assess the chance of success based on historic statistics and analogues of other basins and plays
with similar geological characteristics.
9. 3se sensitivity analysis to determine the critical reservoir parameters.
;. #onduct a probabilistic analysis of reserves using +onte #arlo techniques. If necessary, perform a
further sensitivity analysis here by altering the shapes of the probability distributions assigned to
the reservoir parameters and changing the nature of the dependencies between the variables.
<. A*traction from the probabilistic output of the reserves calculation of some deterministic samples
Jfor e*ample, p5>, p?> and p6> &high, mid, low cases'.
?. 3se sensitivity analysis to determine the critical economic parameters.
@. (erform a probabilistic economic analysis for each deterministic reserve case using +onte #arlo
techniques. If necessary, perform a further sensitivity analysis here by altering the shapes of the
probability distributions assigned to the economic factors and changing the nature of the
dependencies between the variables.
7. 3sing influence diagrams draw the decision tree.
=. "or each reserve case, recombine the chance of success estimated in step 5 and the economic
values generated in step @, through a decision tree analysis to generate A+0s.
6. 3se option theory via decision tree analysis and assess the impact on the A+0.
"igure ?.59: # 9 step #pproach to *n,estment #ppraisal in the =pstream 1il and -as *ndustry
0ariations of the approach could be used for development decisions, any production
decisions and for the decision of when to abandon production and how to
decommission the facilities. 0ersions of it could also be used in other industries with
a similar business environment to the oil and gas industry, for e*ample, the
pharmaceutical or aerospace industries. In these businesses, investment decisions are
similar in scale to the oil industry with the high initial investment without the prospect
of revenues for a significant period and are also characterised by high risk and
uncertainty.
The second research question focussed on two issues. "irst, it aimed to establish
which of the decision analysis techniques that the researcher had identified to
comprise current capability in answering the first research question, upstream oil and
gas companies actually choose to use to make investment decisions. Second, it
5=?
sought to understand how these tools are used in the process of organisational
investment decision2making.
(revious studies into the usage of decision analysis techniques had suggested that
there was a gap between current practice and current capability &for e*ample see
studies by Arnold and Dat1opoulous, 5666F #arr and Tomkins, 566=F Schuyler, 5667F
:uckley et al., 566@ "letcher and .romgoole, 566@F Shao and Shao, 566;F 4im,
"arragher and #rick, 56=<F Stanley and :lock, 56=;F !icks 4elly and (hilippatos,
56=9F :avishi, 56=5F %blak and Delm, 56=> and Stonehill and /athanson, 56@='. It
appeared that whilst the literature described some very sophisticated investment
appraisal tools, companies were choosing to use only the most simplistic. Dowever,
most of the earlier studies had tended to utilise quantitative methodologies and, as
such, these works had only been able to provide an indication of how widely used a
particular decision analysis technique was &for e*ample, Schuyler, 5667'. They had
not provided any insights, based on behavioural decision theory, into the reasons why
some techniques fail to be implemented and others succeed, and, more importantly,
which techniques perform better than others do &#lemen, 5666'. In adopting a
qualitative methodology, the current study was able to address these issues.
Aarlier qualitative research into organisational decision2making had neglected the role
of decision analysis. Several studies had focussed on the e*istence of formalisation
and rationality in decision2making &for e*ample, (apadakis, 566=F .ean and
Sharfman, 566@' but few had e*plicitly e*amined the use, and usefulness, of decision
analysis in investment appraisal. "ewer again had e*amined cases where the decision
situation is characterised by a substantial initial investment, high &absolute' risk and
uncertainty throughout the life of the asset and a long payback period, features that
are common in, though not unique to, the petroleum industry. Typically, where such
research had been undertaken, it had usually been conducted within one company
usually by an employee of that organisation &for e*ample, :urnside, 566=' and had
often not been published due to commercial sensitivity. There had been only one
previous qualitative study researching the use of decision analysis in the oil industry
&"letcher and .romgoole, 566@'. Dowever, this study had only focused on the
perceptions and beliefs of, and decision analysis techniques used by, one functional
area within organisations in the upstream. As such its findings could only be regarded
5=@
as indicative rather than conclusive. In contrast, the current study integrated
perspectives from individuals from a variety of backgrounds within organisations. In
doing so, the research was able to produce a description of current practice in
investment decision2making in the oil industry that was informed from the
perspectives of the main participants in the process.
The current study indicated that for e*ploration decisions, most companies use +onte
#arlo simulation to generate estimates of prospect reserves. They then run their
economic models on only one reserve case. Typically, +onte #arlo simulation is not
used for economic analysis. In production decision2making, the ma8ority of
companies only use deterministic analysis. %ption, portfolio and preference theories
are hardly used at all by any firm. #omparing this approach with the 62step approach
outlined in figure ?.59 and reprinted above, confirms the suggestions of earlier
empirical research, and establishes that there is a gap between current theory and
current practice in the quantitative techniques used for investment appraisal in the
upstream.
The research interviews were then used to provide insights into the reasons for the gap
between current practice and capability. It appears that there are relationships
between decision2makersB perceptions of decision analysis, company culture and the
e*tent to which decision analysis is used for investment appraisal decision2making.
In companies where managers are convinced about decision analysis, the culture is
Hnumbers2drivenK and decision analysis is used e*tensively. In companies where
managers are unconvinced about the value of decision analysis, the company is
largely Hopinion2drivenK and the use of decision analysis is not formalised or
encouraged.
These ideas were then captured in a model of current practice. The *2a*is in the
model relates to the number of decision analysis techniques used for investment
appraisal decisions. The y2a*is of the model indicates the proportion of investment
decisions that are made in each company using decision analysis techniques. (lotting
the interviewed companies on the two a*es then produces the model shown in figure
@.5 and reprinted here. The pattern obtained suggests that organisations begin to use
decision analysis techniques on routine, operational decisions before introducing the
5=7
techniques corporation2wide. Secondly, it provides evidence that as companies
introduce more techniques, they tend to use the techniques on more decisions, some
of which can be regarded as strategic. Thirdly, in the model there are clearly three
groups of companies &each group is a different colour in the figure'. This suggests
that organisations are choosing not to modify which techniques they use or how they
use them, preferring instead to stay within their group. (ossible reasons for this
include the decision2makersB perception of decision analysis, which are affected by
the lack of any empirical evidence to indicate that using decision analysis is
associated with good organisational performance.
"igure @.5 : # model of current practice
This leads into the third research question that this thesis aimed to address, which was
to establish if there is a link between the techniques organisations use for investment
appraisal and good decision2making in the upstream oil and gas industry. This
question was motivated by the recognition that despite over four decades of research
undertaken on developing decision analysis methods, on understanding the
behavioural aspects of decision2making, and on the application of decision analysis in
practice, no previous research had been able to show conclusively what works and
what does not &#lemen, 5666'. Some studies in behavioural decision theory had
evaluated the effectiveness of individual decision analysis techniques &for e*ample,
'roportion o"
decisions
-um#er o" decision analysis tools used
A
1
*
D,)
F,G

I,L,
.
+
M,-,O,',
A
R
0,,
5==
Aldag and (ower, 56=@F ,ohn et al., 56=;F Dumphreys and +c"adden, 56=>' and
#lemen and 4wit &9>>>' had investigated the value of a decision analysis approach in
4odak, but, crucially, no earlier study had shown that use of decision analysis
techniques could actually help organisations to fulfil their ob8ectives.
Gualitative methods again were chosen as the most appropriate to evaluate the
effectiveness, or otherwise, of using a decision analysis approach in organisational
decision2making in the oil industry. 3sing the results from the second stage of the
research, a ranking of the companies according to the number and sophistication of
the techniques and concepts they used, was produced. The research assumed that any
value added to the company from using a decision analysis approach, including HsoftK
benefits, would ultimately affect the bottom2line. This meant that it was possible to
investigate the relationship between the ranking of organisations by their use of
decision analysis generated by the qualitative study and good decision2making
statistically, by using criteria that are indicative of organisational performance in the
upstream. The ma8ority of the results produced suggested that there is a positive
association between the use of decision analysis in investment appraisal and good
organisational performance in the upstream oil and gas industry.
This section has shown how the empirical research presented in this thesis can be used
to answer the three research questions proposed in #hapter 5. The following section
will e*amine how the work produced in this thesis contributes to the e*isting
academic and industry literature. In section =.<, the implications of the study for
practitioners will be investigated.
;(5 ,)OR),I*A+ *O-,RI1U,IO-
This section will demonstrate how the research presented in this thesis can be seen to
have generated a robust set of findings that have contributed to the one of the current
debates in the decision theory literature.
As indicated in #hapter 9, the decision theory literature is comprised of behavioural
decision theory and decision analysis. Simplistically, decision analysis is the label
given to a normative, a*iomatic approach to decision2making under conditions of risk
5=6
and uncertainty. :y using any one, or a combination, of decision analysis techniques,
the decision2maker is provided with an indication of what their decision ought to be
based on logical argument. #onversely, the behavioural decision theory literature
shows that people are not always coherent or internally consistent. They do make
inconsistent patterns of choices and their inferences can be e*ploited &#lemen, 5666',
particularly under conditions of risk and uncertainty.
There is a tendency in the decision theory literature for decision analysts and
behavioural decision theorists to become embroiled in a somewhat circular argument
over the use and benefits of decision analysis. Tocher &567@ and 567= reprinted in
"rench, 56=6' and other behaviouralists argue that people do not behave in the
manner suggested by decision analysis and, in particular, do not adhere to the
underlying assumptions of the decision analysis approach, namely those of rationality
and ma*imising behaviour. Darrison &566? p6>' writes:
Hthe assumptions underlying ma*isimising behaviour are faulty. %b8ectives
are not fi*ed. The known set of alternatives is always incomplete because it is
impossible to obtain perfect information and human beings cognitive
limitations preclude serious consideration of a large number of alternatives.
+any of the variables that must be considered in any attempt at ma*imisation
are not easily quantified. Therefore, a precise preference ranking of the firmBs
ob8ectives or its alternatives that will ma*imise outcome is most unlikely.K
In a special edition in 5665 of the Darvard :usiness $eview )he lo!ic of 'usiness
decision-makin!, At1oni &5665 p<5' commented:
H.ecision2making was never as easy as rationalists would have us think.
(sychologists argue compelingly that even before our present troubles began,
human minds could not handle the comple*ities that important decisions
entailed. %ur brains are too limited. At best, we can focus on eight facts at a
time. %ur ability to calculate probabilities, especially to combine two or more
probabilities J essential for most decision2making J is low +oreover, we
are all prone to let our emotions get in the way J fear for one. Since all
decisions entail risks, decision2making almost inevitably evokes an*iety.K
Additional limitations on ma*imising behaviour become apparent in considering the
human predicament of decision2making. Shackle &567< p5' eloquently articulates this
in the following quote:
56>
HIf choice is originative, it can be effective, it can give thrust to the course of
things intended to secure its ends. In order to secure its ends, choice must
apply a knowledge of what will be the consequence of what. :ut the sequel of
an action chosen by one man will be shaped by circumstance, and its
circumstances will include the actions chosen now and actions to be chosen in
time to come by other men. If, therefore, choice is effective, it is
unpredictable and thus defeats, in some degree, the power of choice itself to
secure e*act ends. This is the human predicament.ecision is not, in its
ultimate nature, calculation, but origination.K
If, as Shackle indicates, decision2making is not founded on calculation, the
assumptions underlying decision analysis are untenable. As such Tocher and other
opponents would rather operate with no model at all than utilise a model that is in
conflict with how people actually act and think:
Hany theory which is worth using predicts how people will behave, not how
they should, so we can do our mathematics.K &Tocher, 567@ reprinted in
"rench, 56=6 p5<>'
.ecision analysts response to such criticisms is that they acknowledge that utility
functions and sub8ective probability distributions do not provide valid models of
decision2makerBs actual preferences and beliefs &"rench, 56=6'. They argue that their
intention is not to describe the decision2makerBs beliefs and preferences as they areF it
is to suggest what they ought to be, if the decision2maker wishes to be consistent.
"rench urges that the HisK should not be confused with the HoughtK, and decision
analysis only suggests how people ought to choose. .ecision analysis, he argues, is
normative not descriptive analysis. 4eeney and $aiffa &567@ pvii' adopt a similar
stance:
HLdecision analysis is aM prescriptive approach designed for normally
intelligent people who want to think hard and systematically about important
real problems.K
4rant1 et al. &5675' describe decision analysis to be:
Hnormative principles defining the concept of rational behaviour rather than
a description of actual behaviourK. &cited by Tocher, 567= reprinted in "rench,
56=6 p5?5'
They go on to say:
H!e want to stress that sub8ective probabilities are means of describing
rational behaviour. /othing moreZ They cannot be used as estimates of the
565
ob8ective probability of an event or the credibility of a statement or the
corroboration of a theory.K &cited by Tocher, 567= reprinted in "rench, 56=6
p5?5'
:ut to critics such as Tocher such a defence is weak and referring to the above quote
from 4rant1, Tocher writes:
HThis sums up my attitude to the utilitariansF I am irritated by their arrogance J
they will tell me how I ought to think regardless of the evidence of how people
actually think or take decisions.K &Tocher, 567= reprinted in "rench, 56=6
p5?5'
"igure =.5 shows the relationships between these two areas of the decision theory
literature. $ecently, researchers have realised that to unite the two seemingly
diametrically opposite views, empirical research needs to establish if there is a
relationship between the use of decision analysis and successful decision2making
&#lemen and 4wit, 9>>>F #lemen, 5666'. #lemen &5666 p9' believes that:
Hsuch research could connect the e*isting areas into a truly unified body
of literatureK
"igure =.5: )he relationship 'et+een decision analysis and 'eha,ioural decision theory (adapted from
Clemen, 1999
Dowever, as indicated above in section =.9, such studies have been slow to appear,
doubtless because of the threat they represent to decision analysts:
HAsking whether decision analysis works is risky. !hat if the answer is
negativeI The contribution will clearly be scientifically valuable, but many
individuals J consultants, academics, instructors J with a vested interest in
decision analysis could lose standing clients, or even 8obs.K &#lemen, 5666
pp9;29<'
Decision analysis
Dow we should decide
1ehavioural
decision theory
Dow we do decide
AR) D)*I0IO-
A-A+C0I0
,)*-IAU)0
)FF)*,I?)B
569
The research presented in this thesis then can clearly be seen to provide a useful
contribution to the theoretical debate, by establishing the e*istence of such an
association in the upstream oil and gas industry.
;(7 IM'+I*A,IO-0 OF ,) 0,UDC FOR 'RA*,I,IO-)R0
The findings presented in this thesis clearly have implications for practitioners in the
oil industry. These will be analysed in this section.
:y answering the first research question, the study has provided an indication of
which decision analysis techniques are the most appropriate for upstream companies
to use for investment appraisal and indicated how these tools can be used together.
#ompanies can use this as a template to modify their own investment appraisal
approach. The model of current practice produced by answering the second research
question, showed which tools companies in the upstream use. This will allow
organisations to compare their processes with the rest of the industry and make any
appropriate modifications. This model also permitted the researcher to identify best
practices in those companies that currently use decision analysis &summarised in
figure =.9' and these should be communicated to companies through publications in
industry 8ournals such as Gournal of %etroleum )echnolo!y.

The decision analysis approach used by the company is formalised. %ften manuals are available to
employees. The manuals detail how the limitations and gaps in the techniques &for e*ample, the
distribution shapes to be used in +onte #arlo simulation' are to be overcome.
.ecision analysis software available throughout the organisation.
Amployees know the decision policy used by the company.
%rganisations have consistent definitions of risk and uncertainty.
All employees have the ability to understand probabilities and communicate probabilistically.
)ood communication between the departments compiling the analysis.
+otivation to conduct decision analysis is high.
.ecision analysis perceived to be a useful tool for quantifying risk and uncertainty.
Aach prospect is sub8ected to peer2review.
.ecision analysis is part of the organisationBs culture.
Amployees trust the results of the analysis.
Avery employee is required to attend training in decision analysis.
+anagement are committed to decision analysis.
+anagement are involved in generating the analysis.
"igure =.9: .est practices in or!anisations7 use of decision analysis
56;
:y providing evidence that decision analysis contributes positively to organisational
performance, the current study ought to promote interest in decision analysis tools and
concepts. This should result in more organisations using decision analysis, and some
companies using the more sophisticated decision analysis techniques and ideas.
#learly then the research presented in this thesis ought to be seen as a vehicle for
narrowing the gap between current practice and current capability in the use of
decision analysis by the upstream oil and gas industry. Dowever, this will only occur
if, simultaneously, decision2makers recognise that decision analysis is not a threat,
that it does not dictate answers, nor does it usurp decision2makers and remove choice
and neither could it ever aspire to do so. /umerous decision analysts &for e*ample,
"rench, 56=6F 4eeney and $affia, 567@' stress that decision analysis is not a means
whereby the decision2maker is replaced by an automatic procedure. /ewendorp
&566@ p7' observes in his book on decision analysis that:
H!e will unfortunately not be able to develop a single Hhandy2dandyK formula
which will cure all the evaluation problems relating to capital investment
decisions.K
The basic presumption of decision analysis is not to replace the decision2makerBs
intuition, to relieve him or her of the obligations in facing the problem, or to be, worst
of all, a competitor to the decision2makerBs personal style of analysis, but to
complement, augment, and generally work alongside the decision2maker in
e*emplifying the nature of the problem &:unn, 56=< p='. 4eeney &56=9' commented
H.ecision analysis will not solve a decision problem, nor is it intended to. Its
purpose is to produce insight and promote creativity to help decision2makers
make better decisions.K &)oodwin and !right, 5665 p<'
Cet, currently decision2makers, in the upstream, at least, appear to fear that
implementation of decision analysis, will be accompanied by a diminishing role for
decision2makers. #learly, decision2makers need to be educated in the conception of
decision2analysis. %nly then will organisations fully adopt decision analysis.
;(8 FU,UR) R)0)AR*
!hilst conducting the research underpinning this thesis, one of the most difficult
tasks for the researcher was to recognise that every interesting issue uncovered could
56<
not be e*plored. As such, whilst contributing to the theoretical debate and providing
useful advice to practitioners, the current study has also highlighted several areas for
future research. These will be discussed in this section.
"irstly, as highlighted in #hapter ?, there is a need for several studies to investigate
the issues surrounding +onte #arlo analysis. %ne study needs to establish the shape
of the input distributions that ought to be used to represent the reservoir parameters, in
a field of specified lithology and depth, in a +onte #arlo simulation to generate an
estimate of the recoverable reserves. A further study is required to e*plore the nature
of the dependencies between these variables. The data necessary for such a study is
due to be published ne*t year by the )eological Society in a book titled 1il and -as
fields of the =nited >in!dom Continental Shelf edited by ,on )luyas et al.. The need
for these studies is particularly pertinent since most companies are using +onte #arlo
analysis to generate estimates of recoverable reserves at the prospect level. Similar
studies also need to be conducted to investigate these issues for economic variables.
Dowever, the economic data that are necessary for such research are regarded by most
companies to be strictly confidential. #onsequently, such research is unlikely to be
undertaken in the near future.
"urther work is also needed to understand the comple*ities of option theory and its
application to the upstream. The growing interest from the industry should ensure
that this occurs. The researcher e*pects to see more companies using the technique on
individual investment appraisal cases in the ne*t couple of years. Software
companies such as +erak are interested in integrating the technique into their e*isting
packages and this should aid its introduction to the industry.
%ne of the most interesting areas that the researcher had to acknowledge was beyond
the scope of this thesis were issues of tacit knowledge and the e*tent to which
organisations and decisions are dependent upon it, and decision2makers reliance on
gut feelings and e*perience. "irstly, understanding such issues, and secondly,
researching them, requires specialist skills in areas, for e*ample, such as
organisational psychology. "ollowing similar observations, in +arch 5666 the
.epartments of +anagement Studies and Aconomics at the 3niversity of Aberdeen
56?
identified a researcher with the appropriate background to undertake such research.
This (h... is due for completion in +arch 9>>9.
"uture research should concentrate on further e*amination of the link between use of
decision analysis and organisational performance. The current study focussed on
those oil companies active on the 34#S, a comparative study could be undertaken in
companies active in other areas. "ollowing presentation of a paper based on #hapter
7 of this thesis at a recent Society of (etroleum Angineers conference in ,apan, ,/%#
&,apanese /ational %il #ompany' are considering conducting a similar study in ,apan.
The study could also be replicated in other industries with a similar high riskNhigh
reward environment, such as pharmaceuticals or aeronautics. #SI$% are currently
considering funding such research.
These studies could perhaps adopt longitudinal research designs. (revious research
&for e*ample, (apadakis and -ioukas, 566@F $a8agopalan et al., 566;', suggests that
organisational performance is a function of a diverse collection of factors. #ause2
effect relationships are, at best, tenuous and a broader conceptualisation of
effectiveness that incorporates both process and performance measures, is now
appropriate &)oll and $asheed, 5667'. 3sing longitudinal research designs,
researchers would be able to gain a greater understanding of the causal relationships
between the decision process and organisational performance by studying how
connections between conte*t, process and outcome unfold over time &(apadakis,
566='. This would minimise the possibility of reverse causality among the main
variables &0an de 0en, 5669F -eonard2:arton, 566>'. #onsequently, longitudinal
research methods would increase researchersB confidence in the causal interpretation
of the findings &Dart and :anbury, 566<F #hakravarthy and .o1, 5669'.
;(: *O-*+U0IO-
This thesis has highlighted that decision analysis should not be perceived to be
providing a dictatorial strait8acket of rationality &"rench, 56=6'. $ather it should be
seen to be a delicate, interactive, e*ploratory tool which seeks to introduce intuitive
8udgements and feelings directly into the formal analysis of a decision problem
&$aiffa, 56@='. The decision analysis approach is distinctive because for each
56@
decision, it requires inputs such as e*ecutive 8udgement, e*perience and attitudes,
along with the Hhard dataK. It helps decision2makers to tread the fine line between ill2
conceived and arbitrary investment decisions made without systematic study and
reflection &He*tinction by instinctK' and a retreat into abstraction and conservatism
that relies obsessively on numbers &Hparalysis by analysisK' &-angley, 566?'. The
thesis has demonstrated that such an approach contributes positively to organisational
performance in the upstream oil and gas industry.
567
A'')-DI*)0
56=
A'')-DII 34 I-,)R?I)/ 0*)DU+)
Roles and responsi#ilities
Assure respondents that all answers will be treated as confidential, that information
disclosed in the interview will be kept securely at the university and that in the final
research report, the real names of people and companies will not be used.
1. !hat is your 8ob titleI &+ake note of company'
9. Typically, what decisions does this mean you are responsible for makingI
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
;. .o you think that two decision2making processes e*ist in organisationsI
Ces /o .onBt know
If so, ask for more e*planation and e*amples.
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
a' In your company what percentage of decisions are regarded as
operationalNproceduralI !hat are the characteristics of an operational decisionI
>29>E
9>2<>E
<>2@>E
@>2=>E
=>25>>E
566
b) !hat type of decisions in your organisation are classified as operational and
whyI Dow are operational decisions madeI !hat is the inputI !ho is involvedI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
c ' !hat type of decisions in your organisation are classified as strategic and whyI
Dow are strategic decisions madeI !ho is involvedI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
Risk and uncertainty
<. !hat do you mean by riskI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
?. !hat do you mean by uncertaintyI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
@. Are these interpretations consistent with company2wide definitions of these termsI
Ces /o .onBt know
If not, what are the company interpretations of risk and uncertaintyI
RI0.
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
U-*)R,AI-,C
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
9>>
7. Are there different interpretations of these terms depending on functional
department, 8ob title etcI
Ces /o .onBt know
If so, ask for details on differences:
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
8.
a' .o you think it is necessary for companies to develop specific training to help
individuals and teams discuss risk and uncertaintyI
Ces /o
b' Das your company developed any such training or vocabularyI
Ces /o .onBt know
Aither way, ask for more details:
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
6.
a' .o you think it is necessary for organisations to develop definitions of the risks it
facesI
Ces /o
b) Das your organisation developed such definitionsI
Ces /o .onBt know
If so, ask what are they are and how these were communicated to employeesI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
9>5
5>. !hat decision2making criteria are used in your companyI
/(0 & /et present value '
A+0 & A*pected monetary value '
.(I & .iscounted (rofit to Investment ratio '
I$$ & Internal rate of return '
$%$ & $ate of return '
.onBt know
%ther & give details below '
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
55.In your company, are pro8ects ranked in terms of risk and rewardI
Ces /o .onBt know
59. Das the ability to better understand and manage risk and uncertainty allowed your
organisation to assume higher levels of riskI
Ces /o .onBt know
5;. Are pro8ects able to be stopped based on assessment of the risks and rewardsI
Ces /o .onBt know
5<. Das your company done anything specific in hiring personnel, or developed
reward and recognition practices to encourage consideration of risk and uncertainty in
decision2makingI
Ces /o .onBt know
9>9
Decision-making techni@ues and so"t2are tools
5?. .oes your organisation use specific tools and processes to analyse and understand
risk and uncertainty in decision2makingI
Ces /o .onBt know
If so, what tools do you use, what stageNlevel in the decision2making process are they
used and how effective are they in this positionI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
5@. !hich, if any, of the following techniques are you aware ofI
+onte2#arlo
.ecision Trees
Scenario (lanning
(ortfolio Theory
,uniper
%ption theory
Analytic hierarchy process
(reference theory
/one
57. Are you aware of any other techniques not listed aboveI
Ces /o
If so, ask for further details:
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
9>;
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
5=. !hat is your opinion of these techniquesI $ank them on a > to ? scale.
&> is least useful and ? is the most useful'
increasing e""ectiveness and e""iciency
+onte #arlo > 5 9 ; < ?
.ecision Trees > 5 9 ; < ?
Scenario (lanning > 5 9 ; < ?
(ortfolio Theory > 5 9 ; < ?
,uniper > 5 9 ; < ?
%ption Theory > 5 9 ; < ?
Analytic hierarchy process > 5 9 ; < ?
(reference theory > 5 9 ; < ?
56. Is any software used to aid decision2making in your organisationI
Ces /o .onBt know
If so,
a' which software packageNs isNare used and whenI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
b' Dow isNare the results of the software toolNs integrated into decision2makingI
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
c' .o you think that thisNthese toolNs have increased the efficiency of the decision2
making procedures in your organisationI
Ces /o
9>. .o you think software tools, in general, assist decision2makingI
9><
Ces /o
95. Are decisions based primarily, on the figures output from the software packagesI
Ces /o .onBt know
99. Dow effective do you think your companyBs decision2making processes and
procedures are in comparison to your main competitorsI
Increasing e""ectiveness and e""iciency
> 5 9 ; < ?
9;. Are any changes likely in the foreseeable future to your current practiceI
Ces /o .onBt know
If so, ask for further details:
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
,hank the intervie2ee "or time taken to conduct intervie2( -ote any comments(
#omments:
9>?
A'')-DII %4 'R)0)-,A,IO-0 A-D 'A')R0
The presentations and papers that have been generated during the current study are
shown below. All were produced before the researcherBs marriage and hence use the
researcherBs maiden name of -amb.
'R)0)-,A,IO-0
5. Methods "or andling Risk and Uncertainty in 1usiness Decision-making,
.epartment of +anagement Studies Seminar, .ecember 5667
9. Methods "or andling Risk and Uncertainty in the Upstream Oil and Gas
Industry, 3ncertainty "orum, 3niversity of Aberdeen, .ecember 5667
;. *urrent *apa#ility in the methods "or handling risk and uncertainty in the
upstream oil and gas industry, presented to 3niversity of Aberdeen, .epartment
of +anagement Studies +astersB students, +arch 566=
<. 'h(D( 'rogress 'resentation, presented to $obert ,ohnston &a representative of
#SI$% J the (h... sponsors', %ctober 566=
?. *urrent *apa#ility versus *urrent 'ractice in decision-making in the
upstream oil and gas industry, presented at the Aconomics and +anagement
Studies %il and )as Seminar, 3niversity of Aberdeen, ,anuary 5666
@. Research Methods, presented to 3niversity of Aberdeen, .epartment of
+anagement Studies +astersB students, +arch 5666
7. 'h(D( 'rogress 'resentation, presented at the (ostgraduate and $esearch
+ethods "orum, 3niversity of Aberdeen, +ay 5666
=. *urrent 'ractice in decision-making in the upstream oil and gas industry,
presented to #onocoBs International $isk and 3ncertainty group, ,une 5666
6. Methods "or handling risk and uncertainty in mature "ields, presented at the
Society of (etroleum Angineers Seminar in Aberdeen, September 5666
5>. Using decision analysis in the upstream oil and gas industry, presented to the
.epartment of +anagement Studies, .ecember 5666
55. Decision analysis and organisational per"ormance4 ,he link 2ith success,
presented at the Aconomics and +anagement Studies %il and )as Seminar,
3niversity of Aberdeen, "ebruary 9>>>
9>@
59. Decision analysis and organisational per"ormance4 ,he link 2ith success,
presented to ,ohn !ils director of 34%%A, +arch 9>>>
5;. 'h(D( 'rogress 'resentation, presented to #SI$% representatives, +arch 9>>>
5<. Decision analysis and organisational per"ormance4 ,he "uture, presented to
#SI$% employees in +elbourne, April 9>>>
5?. ,he Application o" 'ro#a#ilistic and Aualitative Methods to Asset
Management Decision-making, presented at the Society of (etroleum Angineers
Asia (acific #onference on Integrated +odelling for Asset +anagement held in
Cokohama, ,apan, April 9>>>.
'A')R0
0impson, G(0(, Finch, L(( and +am#, F()(, 5666, $isk, uncertainty and relations
between HstrategicK and Hwithin strategyK decision2making in the upstream oil and gas
industry, 3niversity of Aberdeen, .epartment of Aconomics, .iscussion (aper 6625
+am#, F()(, 0impson, G(0( and Finch, L((, 5666, +ethods for evaluating the worth
of reserves in the upstream oil and gas industry, -eopolitics of &ner!y, Issue 99,
/umber <, pp927
0impson, G(0(, +am#, F()(, Finch, L(( and Dinnie, -(*(, 5666, The application of
probabilistic and qualitative methods to decision2making in mature fields, presented at
the Society of (etroleum Angineers conference on the +anagement of +ature "ields
in Aberdeen, September 5666
Finch, L((, Dinnie, -(*(, 0impson, G(0( and +am#, F()(, 5666, A behavioral
framework for understanding decision2making as an inimitable organisational
conference, presented at the AAA(A conference in (rague, /ovember 5666
0impson, G(0(, +am#, F()(, Finch, L(( and Dinnie, -(*(, 9>>>, The application of
probabilistic and qualitative methods to asset management decision2making, paper
submitted to ,ournal of (etroleum Technology.
9>7
A'')-DII 54 0')ARMA- ,)0, FOR *ORR)+A,IO-
The SpearmanBs correlation coefficient is frequently used as alternative to the standard
&(earson' correlation coefficient when only ordinal data are available. SpearmanBs
rank order correlation coefficient is mathematically equivalent to the (earson
correlation coefficient computed on ranks instead of scores. "or the Spearman test,
the scores on each variable must first be rank ordered with the lowest score being
assigned a rank of 5. The SpearmanBs rank order test assumes both variables are
continuous and ordinal and that the data are independent. The procedure usually
followed is:
5. Select the significance level, , and decide whether a one or two tailed test is
required.
9. $ank2order the scores on each variables separately and subtract each ranking on
the response variable from its associated ranking on the e*planatory variable.
#ompute ., the sum of the squares of the resulting differences. "ind n, the
number of scores on either variable.
;. #ompute:
r
s
T5 2 @ .
9
n&n
9
25'
<. If n is larger than ;>, use the normal appro*imation.
?. "rom the table in Appendi* ?, find the relevant critical value&s' of r
s
and make
decision as specified at the top of the table in Appendi* ?
Adapted from -each, 5676.
9>=
A'')-DII 7 P ,) .RU0.A+ /A++I0 A-D /I+*OIO- RA-. 0UM ,)0,0
5. 4$3S4A- !A--IS TAST
The 4ruskal !allis test is a direct generalisation of the !ilco*on $ank Sum Test to
three, or more independent samples. The test attempts to decide whether the samples
of scores come from the same population &the null hypothesis' or from several
populations that differ in location &the alternative hypothesis'. It assumes that the data
are independent and that the scores on the response variable consist of continuous
ordinal data. The procedure usually adopted is:
5. Select the significance level,
9. "ind n, the total number of scores, and t
i
, the number of scores in the ith sample.
#heck that t
i
Tn.
;. $ank order all n scores
<. "ind the sum of ranks in each sample. .enote the rank2sum of the ith sample $
i
.
#heck that $
i
Tn&nU5'N9
?. #alculate:
4T2;&nU5'U 59 $
i
9
n&nU5' t
i
@. If more than a quarter of the response scores are involved in ties, go to 6
7. If more than three samples are being compared or if any of the sample si1es is
larger than ?, use the chi2square appro*imation in 5>.
=. "ind the critical value in the table in Appendi* @. $e8ect the null hypothesis if the
obtained 4 is larger than or equal to the critical value. If the null hypothesis is
re8ected, use multiple comparisons to locate the effects. %therwise, stop.
6. "or e*tensive ties, find u
i
, the number of scores at each particular value of the
response variable and divide the value of 4 obtained in ? by,
5 2 u
i&
u
i
25'& u
i
U5'
n
;
2n
5>. #hi2square appro*imation. !here there are more than two samples, a distribution
known as the chi2square distribution frequently plays a role similar to that of the
normal distribution in two sampled cases in providing an appro*imation to the
9>6
null distribution of a test statistic. This is the case with 4. The table in Appendi*
7 gives critical values of the #hi2square distribution that may be used with sample
si1es beyond the scope of the e*act table in Appendi* @. To obtain the relevant
critical value:
"ind the number of degrees of freedom, the number of samples minus 5.
"ind the relevant critical value in the table in Appendi* 7. $e8ect the null
hypothesis if the obtained 4 &corrected for ties as in 6 if necessary' is larger than or
equal to the critical value. If the null hypothesis is re8ected, use multiple
comparisons to locate the effects. %therwise, stop.
Adapted from -each &5676'
9. +3-TI(-A #%+(A$IS%/S 3SI/) TDA !I-#%W%/ $A/4 S3+ TAST
!hen a significant result has been obtained in the 4ruskal !allis test, all that is
known is that there is some difference in location between the samples. The locus of
this difference is unknown. The !ilco*on $ank Sum test is frequently used for this
purpose and the procedure generally followed is:
5. Select the per e*periment significance level,
9. .ecide on c, the number of comparisons you wish to make. /ormally you will
wish to compare all possible pairs of samples. If there are k samples, the number
of pairs will be cTk&k25'
;. %rder the k samples with respect to their average ranks &given by $
i
Nt
i
' and write
the sample names in order.
<. 3sing a two2tailed test carry out the !ilco*on $ank Sum tests as follows. 3sing
the ordering given in < and the table in Appendi* =, compare the left2most sample
first with the right most, ne*t with the second from the right, and so on until a
non2significant result is obtained. !hen this happens, 8oin the two sample names
with a line. Then take the second sample from the left and compare it first with
the right most, ne*t with the second from the right, and so on until either a non2
significant result is obtained or two samples are being compared that are already
8oined by a line. #ontinue in this way until all comparisons have been e*hausted.
95>
Adapted from -each &5676'
;. TDA 4$3S4A- !A--IS A/. !I-#%W%/ $A/4 S3+ TASTS "%$ ($
5. Arrange the companies into groups depending on their ranking position in the
sophistication of decision analysis rank. &"or e*ample, column 5`T.A`T?
contains companies that were ranked in the top ? in the decision analysis ranking'.
Then note companies proved reserves under the appropriate heading. The
following simple notation will be used, n will be the total number of companies, t
5
will be the number in 5`T.A`T?, t
9
will be the number in @`T.A`T5>, t
;
will be
the number in 55`T.A`T5<.
9. $ank the data from low to high &with rank of 5 being assigned to the company
with the least proved reserves'. #ompute the rank sums for the three samples.
This is done in the table below, where $
i
refers to the rank sum of the ith group, so
$
5
T?7, $
9
T;< and $
;
T5<. The test statistic will need to take account of what is
going on in all ; groups to give an adequate picture of the data. Since there are
different numbers of companies in some of the groups, the mean rank in each
group is more informative than the rank sum. This is obtained by dividing the
rank sum $
i
for a given group by the number of companies in that group and is
shown in the table below for each group. If the null hypothesis is true, then these
three average ranks should all be relatively close together. In this case the average
ranks differ, this indicates the alternative hypothesis is true.
3HMDAHM8 :HMDAHM3& 33HMDAHM37
5>
6
5;
55
5<
?
59
@
;
=
<
9
7
5
t3M 8, R3M89, RDt M 33(7 t%M8, R%M57, RDtM:(; t5M7, R5M37, RDtM5(8
;. The test statistic needs to reflect how different the average ranks are, that is it
should be a measure of the dispersion of the ranks. -each provides the proof of
the formula &5676 pp5<625?>', here it will be take as given:
955
4T 2;&nU5'U 59 $
i
9
n&nU5' t
i
<. In this case, 4T=.5<9=
There are 9 degrees of freedom in this case.
It is therefore possible to re8ect the null hypothesis at T>.>?
?. To find the locus of the difference, use the rank sum test and choose T>.>9 as the
significance level.
@. !riting the groups in order of increasing $NT:
33HMDAHM37 :HMDAHM3& 3HMDAHM8
<
9
7
5
?
59
@
;
=
5>
6
5;
55
5<
t3M 8, R3M89, RDt M 33(7 t%M8, R%M57, RDtM:(; t5M7, R5M37, RDtM5(8
7. Then comparing 55`T.A`T5< with 5`T.A`T? gives:
pNs @Ns
?
?
?
?
>
>
>
>
'M%& AM&
ST(2GT9>, which is significant at T>.>9
=. Then comparing 55`T.A`T5< with @`T.A`T5> gives:
959
pNs @Ns
<
?
;
?
5
>
9
>
(T57 GT;
ST(2GT5<, which is not significant at T>.>9.
6. So the result of the !ilco*on $ank Sum for ($ is, 55`T.A`T5<, @`T.A`T5>,
5`T.A`T?
95;
A'')-DII 84*RI,I*A+ ?A+U)0 OF FOR 0')ARMA- ,)0,0
&0amples: "or a two tailed test, with nT5? and T>.>?, re8ect the null hypothesis if
the obtained is larger than or equal to >.?95, or if it is smaller than or equal to
2>.?95.
&0amples: "or a one tailed test, with nT5? and T>.>?, if an upper tail test is required,
re8ect the null hypothesis if the obtained is larger than or equal to >.<<@. If a lower
tail test is required, re8ect if the obtained is smaller than or equal to 2>.<<@.
1ne tailed si!nificance le,el,
>.5 >.>? >.>9? >.>5 >.>>? >.>>5
)+o tailed si!nificance le,el,
n >.9 >.5 >.>? >.>9 >.>5 >.>>9
< 5 5
? >.= >.6 5 5
@ >.@?7 >.=96 >.=@@ >.6<; 5.>>
7 >.?75 >.75< >.7=@ >.=6; >.696 5.>>
= >.?9< >.@<; >.7;= >.=;; >.==5 >.6?9
6 >.<=; >.@ >.7 >.7=; >.=;; >.657
5> >.<?? >.?@< >.@<= >.7<? >.76< >.=76
55 >.<97 >.?;@ >.@5= >.7>6 >.7?? >.=<?
59 >.<>@ >.?>; >.?=7 >.@7= >.797 >.=5=
5; >.;=? >.<=< >.?@> >.@<= >.7>; >.765
5< >.;@7 >.<@< >.?;= >.@9@ >.@76 >.775
5? >.;?< >.<<@ >.?95 >.@>< >.@?7 >.7?>
5@ >.;<5 >.<96 >.?>; >.?=? >.@;? >.796
57 >.;9= >.<5< >.<== >.?@@ >.@5= >.755
5= >.;57 >.<>5 >.<7< >.??> >.@ >.@69
56 >.;>6 >.;65 >.<@ >.?;? >.?=< >.@7<
9> >.966 >.;= >.<<7 >.?99 >.?7 >.@@
"rom -each &5676'
95<
A'')-DII :4*RI,I*A+ ?A+U)0 OF . FOR .RU0.A+ /A++I0 ,)0, /I, 5
I-D)')-D)-, 0AM'+)0
t
5
is the number of observations in the largest sample
t
;
is the number of observations in the smallest sample
&0ample: "or t
5
T<, t
9
T;, t
;
T9, and T>.>?, re8ect the null hypothesis if the obtained 4
is larger than or equal to ?.;;
The table may be used for any case for which t
5
, t
9
, t
;
are all less than si*. The test is
inherently two tailed.
Si!nificance le,el,
t
1
t
5
t
6
>.9 >.5 >.>? >.>9? >.>5 >.>>? >.>>5
9 9 5 ;.@
9 9 9 ;.75 <.?7
; 9 5 ;.?9 <.96
; 9 9 ;.6; <.? <.75
; ; 5 ;.96 <.?7 ?.59
; ; 9 ;.7= <.?@ ?.;@ ?.?@
; ; ; ;.<7 <.@9 ?.@ ?.6@ 7.9 7.9
< 5 5 ;.?7
< 9 5 ;.5@ <.?
< 9 9 ;.@7 <.<@ ?.;; ?.?
< ; 5 ;.95 <.>@ ?.95 ?.=;
< ; 9 ;.<< <.?5 ?.<< @ @.<< 7
< ; ; ;.;6 <.75 ?.7; @.5? @.7? 7.;9 =.>9
< < 5 ;.97 <.57 <.67 @.57 @.@7
< < 9 ;.<@ <.?? ?.<? @.>= 7.>< 7.9=
< < ; ;.<9 <.?? ?.@ @.;6 7.5< 7.@ =.;;
< < < ;.? <.@? ?.@6 @.@9 7.@? = =.@?
? 5 5 ;.=@
? 9 5 ;.;; <.9 ?
? 9 9 ;.;@ <.;7 ?.5@ @ @.?;
? ; 5 ;.99 <.>9 <.6@ @.><
? ; 9 ;.<5 <.@? ?.9? @ @.=9 7.56
? ; ; ;.<< <.?; ?.@? @.;9 7.>= 7.?9 =.9<
? < 5 ;.>6 ;.66 <.66 ?.7= @.6? 7.;@
? < 9 ;.;@ <.?< ?.97 @.>< 7.59 7.?7 =.55
? < ; ;.;9 <.?? ?.@; @.<5 7.<? 7.65 =.?
? < < ;.;; <.@9 ?.@9 @.@7 7.7@ =.5< 6
? ? 5 ;.9< <.55 ?.5; @ 7.;5 .7.7?
? ? 9 ;.;6 <.?5 ?.;< @.;? 7.97 =.5; =.@6
? ? ; ;.<; <.?? ?.75 @.<6 7.?< =.9< 6.>@
? ? < ;.;5 <.?9 ?.@< @.@7 7.77 =.;7 6.;9
? ? ? ;.<9 <.?@ ?.7= @.7< = =.79 6.@=
"rom -each &5676'
95?
A'')-DII 94 *RI,I*A+ ?A+U)0 OF *I-0AUAR) A, ,) &(&8 A-D &(&3 +)?)+ OF
0IG-IFI*A-*)
&0ample: !ith T>.>? and < degrees of freedom re8ect the null hypothesis if the
critical value obtained is larger than or equal to 6.<==. The distribution is
inherently two tailed
Le,el of si!nificance,
df >.>? >.>5
5 ;.=<5 @.;?
9 ?.665 6.95>
; 7.=5? 55.;<?
< 6.<== 5;.977
? 55.>7> 5?.>=@
@ 59.?69 5@.=59
7 5<.>@7 5=.<7?
= 5?.?>7 9>.>6>
6 5@.656 95.@@@
5> 5=.;>7 9;.9>6
55 56.67? 9<.79?
59 95.>9@ 9@.957
5; 99.;@9 97.@==
5< 9;.@=? 96.5<5
5? 9<.66@ ;>.?7=
"rom -evin and "o* &56=='
95@
A'')-DII ;4 *RI,I*A+ ?A+U)0 OF 0 FOR ,) /I+*OIO- RA-. 0UM ,)0,
t
5
is the number of observations in the largest sample
t
9
is the number of observations in the smallest sample
&0ample: "or a two2tailed test, with t
5
T;, t
9
T9 and T>.>?, re8ect the null hypothesis if
the obtained S is larger than or equal to @, or if it is smaller than or equal to J@.
&0ample: "or a one2tailed test, with t
5
T;, t
9
T9 and T>.>?, if an upper tailed test is
required re8ect the null hypothesis if the obtained S is larger than or equal to ?=. If a
lower tail test is required, re8ect if the obtained S is smaller than or equal to J?=.
The table may be used for any case for which t
5
, t
9
, t
;
are all less than si*. The test is
inherently two tailed.
1ne tailed si!nificance le,el,
>.5 >.>? >.>9? >.>5 >.>>? >.>>5
)+o-tailed si!nificance le,el,
t
1
t
5
>.9 >.5 >.>? >.>9 >.>5 >.>>9
; 9 @
; ; 7 6
< 9 =
< ; 5> 59
< < 5> 5< 5@
? 9 = 5>
? ; 55 5; 5?
? < 59 5@ 5= 9>
? ? 5? 57 95 9;
@ 9 5> 59
@ < 5< 5= 9> 99 9<
@ ? 5@ 9> 9< 9@ 9=
@ @ 5= 99 9@ ;> ;9
"rom -each &5676'
957
1I1+IOGRA'C
95=
Aldag, $.,., and (ower, ..,., 56=@, An empirical assessment of computer aided
decision analysis, :ecision Sciences, 57, pp?792?==
Aldrich, D.A., 5676, 1r!anisations and en,ironments, Anglewood #liffs, (rentice Dall,
/ew ,ersey
Allaire, C., and "irsitrotu, +.A., 56=6, #oping with strategic uncertainty, Sloan
2ana!ement <e,ie+, ;, pp725@
Allais, +., 56?;, -e comportement de lBhomme rationnel devant le risque: #ritique des
postulates et a*imoes de lBecole americaine, &conometrica, 95, pp?>;2?<@
Allias, +., and Dagen, ,., editors, 5676, &0pected utility hypotheses and the #llais
parado0, .ordrecht, The /etherlands, $eidel
Anderson, :."., .eane, ..D., Dammond, 4.$., and +c#lelland, ).D., 56=5, Concepts
and Gud!ement in :ecision <esearch, (raeger, /ew Cork
Argote, -., 56=9, Input uncertainty and organisational co2ordination in hospital
emergency units, #dministrati,e Science Huarterly, 97, pp<9>2<;<
Argote, -., Seabrigh, +.A., and .yer, -., 56=@, Individual versus group use of base
rate and individuating information, 1r!anisation .eha,iour and Duman :ecision
%rocesses, ;=, pp@?27?
Armstrong, ,.S., .enniston, !.:., and )ordon, +.+., 567?, The use of decomposition
principle in making 8udgements, 1r!anisation .eha,iour and Duman :ecision
%rocesses, 5<, pp9?729@;
Arnold, ).#., Dat1opoulous, (..., 9>>>, The theory2practice gap in capital budgeting:
Avidence from the 3nited 4ingdom, Gournal of .usiness (inance and #ccountin!,
,une, volume 97, numbers ? and @, pp@>;2@9@
Atrill, (., 9>>>, (inancial 2ana!ement for @on-specialists, Second Adition, (rentice
Dall, Angland
:ailey, :., #ouet, :., -amb. "., $ose, (., and Simpson, )., in press, Taking a
#alculated $isk, 1ilfield <e,ie+
:all, :.#. and Savage, S.-., 5666, Dolistic vs. Dole2istic A^( Strategies, S(A ?77>5,
Gournal of %etroleum )echnolo!y, September, pp7<2=<
:arley, S., 566?, Images of imaging: notes on doing longitudinal fieldwork, in Duber,
). and Duber, ). and 0an de 0en, A. &editors', Lon!itudinal (ield <esearch 2ethods:
Studyin! processes of or!anisational chan!e, Sage, -ondon
:aumard, (., 5666, )acit >no+led!e in 1r!anisations, Sage (ublications, -ondon
956
:avishi, 0.:., 56=5, #apital budgeting practices at multinationals, 2ana!ement
#ccountin!, August, pp;92;?
:a*ter, +. and $ennie, A., 566@, # financial calculus: #n introduction to deri,ati,e
pricin!, #ambridge 3niversity (ress, #ambridge
:echofer, "., 567<, #urrent Approaches to empirical research: some central ideas, in
$e*, ,., &editor', #pproaches to Sociolo!y: an introduction to ma$or trends in .ritish
Sociolo!y, $outledge, 4egan (aul, -ondon
:ell, #. and Ancel, S., 567=, &editors', *nside the 3hale, (eragamon (ress, %*ford
:ell, #. and /ewby, D., 5677, &editors', :oin! Sociolo!ical <esearch, Allen^3nwin,
-ondon
:ernoulli, .., 57;=, Specimen theoriae novae de mensura sortis, translated by -ouise
Sommer published in &conometrica, volume 99, 56?<, pp9;2;@
:ernoulli, ,., 575;, #rs con$ectandi tractatus de seri'us infinitis, :asel
:ernstein, (.-., and Silbert, T.D., 56=<, 4eeping Informed, Dar,ard .usiness <e,ie+,
@9&?', pp;92<>
:lack, ". and Scholes, +., 567;, The pricing of options and corporate liabilities,
Gournal of %olitical &conomy, =5, pp@;72??<
:lair, ,..., Slaton, #.$., and Savage, ).T., 566>, Dospital (hysician ,oint 0entures: a
Strategic Approach for :oth .imensions of Success, Dospital and Dealth Ser,ices
#dministration, ;?&5', pp;29@
:orsch, 4. and +ossin, ,., &editors', 56@=, <isk and =ncertainty, proceedings of a
conference held by the International Aconomic Association, +acmillan, /ew Cork
:ourgeois, -.,., III, 56=?, Strategic goals, perceived uncertainty, and economic
performance in volatile environments, #cademy of 2ana!ement Gournal, 9=, pp?<=2
?7;
:ourgeois, -.,., III, 56=<, Strategic management and determinism, #cademy of
2ana!ement <e,ie+, 6, pp?=@2?6@
:ourgeois, -.,., III, 56=5, %n the +easurement of %rganisational Slack, #cademy of
2ana!ement <e,ie+, @, pp962;6
:ourgeois, -.,., III, and Aisenhardt, 4.+., 56==, Strategic .ecision (rocesses in Digh
0elocity Anvironments: "our cases in the +icrocomputer Industry, 2ana!ement
Science, ;<&7', pp=5@2=;?
:owman, #. and Ambrosini, 0., 5667, 3sing Single $espondents in Strategy $esearch,
.ritish Gournal of 2ana!ement, volume =, pp55625;5
99>
:( statistical review of world energy, 9>>>, http:NNwww.bp.co.uk
:realey, $.A and +yers, S.#., 566@, %rinciples of Corporate (inance, +c)raw2Dill,
/ew Cork
:rennan, +.,., and Schwart1, A.S., 56=?, Avaluating /atural $esource Investments,
Gournal of .usiness, ?=&?', p5;?
:rown, +., 5669 Thinking small about oil, 2ana!ement )oday, +arch, p=>2=<
:runsson, /., 56=?, )he irrational or!anisation, !iley, #hichester
:ryman, A., and :urgess, $.)., 566<, &editors', #nalysin! Hualitati,e :ata,
$outledge, -ondon
:ryman, A., 56=6, <esearch method and or!anisation studies, )eorge 3nwin, -ondon
:uckley, A., 9>>>, 2ultinational (inance, <
th
edition, (rentice Dall, Angland
:uckley, A., :uckley, (., -angevin, (., and Tse, 4., 566@, The financial analysis of
foreign investment decisions by large 34 based companies, &uropean Gournal of
(inance, 9&9', pp5=529>@
:udescu, ..0., and !allsten, T.S., 566?, (rocessing linguistic probabilities, in
:usemeyer, ,. and +edin, ..-., &editors', :ecision-makin! from a co!niti,e
perspecti,e, pp97?2;5=
:unn, .., 56=<, #pplied decision analysis, +c)raw Dill, /ew Cork
:urgess, $.)., 56=<a, )he <esearch %rocess in &ducational Settin!s: )en Case
Studies, "almer (ress, -ondon
:urgess, $.)., 56=<b, *n the (ield: #n introduction to field research, )eorge Allen and
3nwin, -ondon
:urke, -.A., and +iller, +.4., 5666, Taking the mystery out of intuitive decision
making, )he #cademy of 2ana!ement &0ecuti,e, volume 5;, issue <, /ovember, p56
:urnside, #.A., 566=, #pplications of <isk #nalysis and *n,estment #ppraisal
)echni"ues in :ay to :ay operations of an upstream oil company, submitted in partial
fulfillment of the degree of +aster of :usiness Administration, 3niversity of
Aberdeen, unpublished.
:utler, $., 5665, :esi!nin! or!anisations / # decision-makin! perspecti,e, $outledge,
-ondon
#ampbell, #.,., 5667, )he Comin! 1il Crisis, +ulti2Science (ublishing #ompany and
(etroconsultants, S.A.
#apon, /., "arley, ,.3., and Dulbert, ,.+., 566<, Strategic (lanning and "inancial
(erformance: +ore Avidence, Gournal of 2ana!ement Studies, ;5&5', pp5>?255>
995
#arr, #., and Tomkins, #., 566=, #onte*t, culture, and the role of the finance function
in strategic decisions. A comparative analysis of :ritain, )ermany, the 3.S.A. and
,apan, 2ana!ement #ccountin! <esearch, 6, pp95;29;6
#hakravarthy, :.S., and .o1, C., 5669, Strategy process research: "ocussing on
corporate self2renewal, Strate!ic 2ana!ement Gournal, 5;, pp?25<
#hang, $.C., and +organ, +.!., 9>>>, (erformance Scorecards: +easuring the $ight
Things in the $eal !orld, D$ +aga1ine, August, volume <=, issue =, p57>
#hapman, -.,., and #hapman, ,.(., 56@6, Illusory correlation as an obstacle to the use
of valid psychodiagnostic signs, Gournal of #'normal %sycholo!y, 7<, pp97529=>
#hild, ,., 567<, !hat .etermines %rganisation (erformanceI The 3niversals vs. the it2
all2depends, 1r!anisation :ynamics, 5, pp925=
#hild, ,., 5679, %rganisational structure, environment, and performance: the role of
strategic choice, Sociolo!y, @, pp9295
#lemen, $., 5666, :oes :ecision #nalysis 3orkI # <esearch #!enda, "urqua School
of :usiness, unpublished, first presented at a seminar at the 3niversity of Te*as,
Austin, April, 566@
#lemen, $., ,ones, S.4., and !inkler, $.-., 566@, Aggregating "orecasts: An empirical
evaluation of some :ayesian +ethods, in :erry, .., #haloner, 4., and )eweke, ,.,
&editors', .ayesian Statistics and &conometrics: &ssays in Donor of #rnold Jeller,
pp925;, !iley, /ew Cork
#lemen, $., and 4wit, $.#., 9>>>, The 0alue of .ecision Analysis at Aastman 4odak
#ompany, 566>2566=, forthcoming in *nterfaces
#ohen, +. S., 56=6, A database tool to support probabilistic assumption based
reasoning in intelligence analysis, %roceedin!s of the 19A9 Goint :irector of the C
5
Symposium, "t. +c/air, 0A, ,une, pp97296
#ohen, +.S., Schum, ..A., "reeling, A./.S., and #hinnis, ,.%., 56=?, 1n the art and
science of hed!in! a conclusion: # co!niti,e approach, "alls #hurch, .ecision
Science #onsortium
#ooper, +. and :urford, S., 5666, +urchison "ield: Hdoing things betterK, published in
the proceedings of <eser,oir 2ana!ement of 2ature (ields, S(A Aberdeen Section,
96
th
September
#opeland, T., 4oller, T., +urrin, ,., 566>, Caluation: 2easurin! and 2ana!in! the
Calue in Companies, ,ohn !iley and Sons, /ew Cork #ity
999
#orbin, $.+., 56=>, .ecisions that might not get made, in !allsten, T., &editor',
Co!niti,e %rocesses in choice and decision 'eha,iour, Dillsdale, /ew ,ersey, Arlbaum
#yert, $.+., and +arch, ,.)., 56@;, # .eha,ioural )heory of the (irm, (rentice Dall
.as, T.4., )osavi, A., +ahadevan, S., +archalleck, /., 5666, Solving Semi2+arkov
.ecision (roblems 3sing Average $eward $einforcement -earning, 2ana!ement
Science, volume <?, number <, April
.aft, $.-., 56=?, !hy I recommend that your manuscript be re8ected and what you can
do about it, #ummings, -.-., "rost, (.(. &editors', %u'lishin! in the 1r!anisational
Sciences, $ichard .. Irwin, Domewood, Illinois
.atastream, http:NNwww.datastreaminsite.com
.awes, $.+., 56==, <ational Choice in an =ncertain 3orld, Darcourt :race
,ovanovich, San .iego, #alifornia
.ean, ,.!. and Sharfman, +.(., 566@, .oes decision process matterI A study of
strategic decision2making effectiveness, #cademy of 2ana!ement Gournal, April,
volume ;6, number 9, pp;@=267
.e +organ, A., 5=<7, (ormal Lo!ic: )he calculus of inference, necessary and
pro'a'le, -ondon
.en1in, /.4., 567=, Sociolo!ical 2ethods: # Source'ook, +c)raw2Dill, /ew Cork
.epartment of Trade and Industry, 9>>>, => &ner!y in .rief, http:NNwww.dti.gov.uk
.epartment of Trade and Industry, 5666, )he oil and !as industry )ask (orce <eport #
)emplate for Chan!e, pp529
.epartment of Trade and Industry, 566=, )he &ner!y <eport, volume 9, The Stationery
%ffice, -ondon
.ess, ).)., 56=7, #onsensus on Strategy "ormulation and %rganisational
(erformance: #ompetitors in a "ragmented Industry, Strate!ic 2ana!ement Gournal,
=, pp9?62977
.ess, ).). and %riger, /.4., 56=7, Anvironment, Structure and #onsensus in Strategy
"ormulation: A #onceptual Integration, #cademy of 2ana!ement <e,ie+, 59&9',
pp;5;2;;>
.ess, ).)., and $obinson, $.:., 56=<, +easuring %rganisational (erformance in the
Absence of %b8ective +easures: The case of the privately2held firm and conglomerate
business unit, Strate!ic 2ana!ement Gournal, ?, pp9@?297;
99;
.e!oot, (., Deyvaert, D., and +artou, "., 567727=, Strategic +anagement: An
empirical study of 5@= :elgian "irms, *nternational Studies of 2ana!ement and
1r!anisation, 7, pp@>27?
.i*it, A.4. and (indyck, $.S., 566=, The options approach to capital investment, in
$utherford, ,. &editor', (inancial strate!y: #ddin! Stakeholder Calue, !iley,
#hichester
.i*it, A.4., and (indyck, $.S., 566<, *n,estment =nder =ncertainty, (rinceton
3niversity (ress
.rury, #., 56=?, 2ana!ement and Cost #ccountin!, von /ostrand $einhold,
!okingham
Adwards, !., 56?<, The theory of decision2making, %sycholo!ical .ulletin, <5, pp;=>2
<57
Akenberg, -., 9>>>, The logic of conflicts between decision making agents, Gournal of
lo!ic and computation, volume 5>, number <, pp?=;2@>9
Ailon, S. and "awkes, T.$., 567;, Sampling (rocedures for $isk Simulation,
1perational <esearch Huarterly, 9<, number. 9, pp9<529?9
Aisenhardt, 4.+., 56=6, +aking fast decisions in high velocity environments,
#cademy of 2ana!ement Gournal, ;9, pp?<;2?77
Aisenhardt, 4.+. and :ourgeois, -.,., III, 56==, (olitics of strategic decision2making
in high velocity environments: Towards a midrange theory, #cademy of 2ana!ement
Gournal, ;5, pp7;7277>
Anergy .ay, 9>>>, 1il a!e leakin! a+ay, ;
rd
,uly
At1ioni, A., 5665, Dumble .ecision2making, )he lo!ic of 'usiness decision-makin!,
Darvard :usiness $eview
"arquhar, (.D., 56=<, 3tility assessment methods, 2ana!ement Science, ;>, pp59=;2
5;>>
"eldman, +.A., and +arch, ,.)., 56=5, Information in organisations as signal and
symbol, #dministrati,e Science Huarterly, 97, pp57525=@
"erner, A., 56=6, The $esearch (rocess in the #ross /ational Industrial $elations #ase
Study: A personal account, paper ori!inally prepared for a talk to doctoral researchers
at the Centre for #d,anced Studies in the Social Sciences, Insituto de ,uan +arch de
Astudios e Investigaciones, 59
th
April
"inancial Times, http:NNwww.ft.com
99<
"inch, ,.D., and +c+aster, $., 9>>>, 1n non-parametric statistical inference in the
pursuit of causal e0planations, submitted to the #ambridge $ealist !orkshop $eunion
#onference, +ay
.e "inetti, :, 56@<, "oresight: Its logical laws, its sub8ective sources, in 4yberg, A.,
and Smoker, A., &editors', Studies in Su'$ecti,e %ro'a'ility, &essay originally
published, 56;7, !iley'
"ischer, )., 5677, #onvergent validation of decomposed multi2attribute utility
assessment procedures for risky and riskless decisions, 1r!anisational .eha,iour and
Duman %erformance, 5=, pp96?2;5?
"ischer, )., 567?, #n &0perimental Study of (our %rocedures (or #!!re!atin!
Su'$ecti,e %ro'a'ility #ssessments, )echnical <eport 74-7, .ecisions and .esigns,
Inc., !ashington ..#.
"ischoff, :., 56=9, .ebiasing, in 4ahneman, .., Slovic, (., Tversky, A., &editors',
Gud!ement under uncertainty: Deuristics and .iases, pp<992<<<, #ambridge
3niversity (ress, #ambridge
"ischoff, :., 56=>, .ecision Analysis: #linical Art or #linical Science, in S8oberg, -.,
Tys1ka, T., !ise, ,.A., &editors', Duman :ecision-makin!, :odfors, .o*a
"letcher, A.,.(., and .romgoole, (., 566@, Current %ractices of handlin! su'surface
uncertainty, unpublished ,uniper .ocument +$6?>5N>=, %ctober
"ord, #.+. and )ioia, .. A., 9>>>, "actors influencing creativity in the domain of
managerial decision making, Gournal of 2ana!ement, ,uly, volume 9@, number <,
pp7>?27;9
"ord, #.+., 56=6, Creati,ity in mana!erial decision-makin!: #n e0amination of
factors that influence the creati,ity of mana!ers7 decisions, 3npublished doctoral
dissertation, (ennsylvania State 3niversity, 3niversity (ark
"redrickson, ,.!., 56=?, Affects of .ecision +otive and %rganisational (erformance
-evel on Strategic .ecision (rocesses, #cademy of 2ana!ement Gournal, 9=&<',
pp=952=<;
"redrickson, ,.!., 56=<, The #omprehensiveness of Strategic .ecision (rocesses:
A*tension, %bservations, "uture .irections, #cademy of 2ana!ement Gournal, 97&;',
pp<<?2<@@
"redrickson, ,.!., and Iaquinto, A.-., 56=6, Inertia and #reeping $ationality in
Strategic .ecision (rocesses, #cademy of 2ana!ement Gournal, ;9&<', pp?5@2?<9
99?
"redrickson, ,.!., and +itchell, T.$., 56=<, Strategic decision processes:
comprehensiveness and performance in an industry with an unstable environment,
#cademy of 2ana!ement Gournal, 97, pp;662<9;
"rench S., 56=6, <eadin!s in decision analysis, #hapman and Dall -td., -ondon
"rench, S., 56=<, "rom decision theory to decision analysis, in Aglese, $.! and $and,
).4, &editors', :e,elopments in 1perational <esearch, pp772=7, (ergamon (ress,
%*ford
)albraith, ,.$., 567;, :esi!nin! Comple0 1r!anisations, Addison2!esley, $eading,
+ass.
)alli, A.)., Armstrong, +. and ,ehl, :., 5666, #omparison of three methods for
evaluating oil pro8ects, Gournal of %etroleum )echnolo!y, %ctober, pp<<2?>
)ans, ,.S., 5666, -imited information, the possibility of rational choice and contingent
coping valuation method, *nternational Gournal of Social &conomics, ,anuary, volume
9@, number 9;, pp<>92<5<
)igeren1er, )., 5665, Dow to make cognitive illusions disappear: :eyond Hheuristics
and biasesK, &uropean <e,ie+ of Social %sycholo!y, 9, pp=;255?
)igeren1er, ). and Doffrage, 3., 566?, Dow to improve :ayesian $easoning without
instruction: "requency "ormats, %sycholo!ical <e,ie+, 5>9, pp@=<27><
)ill, ,. and ,ohnson, (., 5665, <esearch 2ethods for 2ana!ers, (aul #hapman
(ublishing, -ondon
)luyas, ,.)., Dichens, D., Avans, I.,. and Dogg, A.,.#., &editors', 9>>5, 1il and !as
fields of the =nited >in!dom Continental Shelf, )eological Society, -ondon
)oldstein, !.+., and Ainhorn, D.,., 56=7, A*pression theory and the preference
reversal phenomena, %sycholo!ical <e,ie+, 6<, pp9;@29?<
)oll, I., and $asheed, A.+A., 5667, $ational .ecision2making and "irm (erformance:
The moderating role of environment, Strate!ic 2ana!ement Gournal, 5=&7', pp?=;2?65
)oodwin, (., and !right )., 5665, :ecision analysis for mana!ement $ud!ement, ,ohn
!iley^Sons -imited, #hichester, Angland
)ough, $., 567?, )he &ffects of -roup (ormat on #!!re!ate Su'$ecti,e %ro'a'ility
:istri'utions, =tility, %ro'a'ility and Duman :ecision-makin!, $eidel, .ordrecht,
Dolland
)randori, A., 56=<, A prescriptive contingency view of organisational decision2
making, #dministrati,e Science Huarterly, 96, pp56929>6
99@
)rayson, #.,., 56@>, :ecisions under uncertainty: oil and !as drillin! decisions,
Darvard 3niversity (ress, #ambridge
)rether, ..+., and (lott, #.$., 5676, Aconomic theory of choice and the preference
reversal phenomenon, #merican &conomic <e,ie+, @6, pp@9;2@;=
)rinyer, (.S., +c4iernan and Casai2Ardekani, +., 56==, +arket %rganisational and
+anagerial #orrelates of Aconomic (erformance in the 3.4. Alectrical Angineering
Industry, Strate!ic 2ana!ement Gournal, 6, pp9672;5=
)rinyer, (.S., and /orburn, .., 567727=, (lanning for e*isting markets: an empirical
study, *nternational Studies in 2ana!ement and 1r!anisation, 7, pp662599
)unn, :., 9>>>, .ecisions, .ecisions, .ecisions, Strate!ic (inance, volume =5, issue
7, p5<
)ustafson, ..D., Shukla, .elbecq, A., and !alster, ).!., 567;, A #omparative Study
of .ifferences in Sub8ective -ikelihood Astimates +ade by Individuals, Interacting
)roups, .elphi )roups, and /ominal )roups, 1r!anisational .eha,iour and Duman
%erformance, 6, pp9>>2965
)utleber, ..S., Deigerger, A.+., +orris, T..., 566?, Simulation Analysis for Integrated
Avaluation of Technical and #ommercial $isk, Gournal of %etroleum )echnolo!y,
.ecember, p5>@9
Dalley, A., and )uilhorn, A., 5667, -ogistic behaviour of small enterprises:
performance, strategy and definition, *nternational $ournal of physical distri'ution and
lo!istics mana!ement, ,uly2August, volume 97, numbers 72=, p<7?
Dambrick, ..#. and Snow, #.#., 5677, A conte*tual model of strategic decision2
making in organisations, #cademy of 2ana!ement %roceedin!s, pp5>62559
Dammersley, +., 5669, 3hat7s +ron! +ith ethno!raphyI, $outledge, -ondon
Dammersley, +. and Atkinson, (., 56=;, &thno!raphy: %rinciples in %ractice,
-ongman, -ondon
Dammond, ,., 4eeney, $., $aiffa, D., 5666, Smart Choices, Darvard :usiness School
(ress
Dammond, ,., 56@7, :etter decisions with preference theory, Dar,ard .usiness
<e,ie+, /ovember2.ecember, pp59;25<5
Dammond, (., 56@<, &editor', Sociolo!ists at +ork, :asic :ooks, /ew Cork
Darper, ,., 56=<, 2easurin! .usiness %erformance, Institute of +anpower Studies
Series /o. 7, )ower (ublishing #ompany -td., Angland
997
Darrison, A. "., 566?, )he mana!erial decision-makin! process, Doughton2+ifflin,
3SA
Darrison, A."., and (elleteir, +.A., 9>>>, -evels of strategic decision success,
2ana!ement :ecision, 96
th
"ebruary, volume ;=, number 9, pp5>7255=
Dart, S.-., 5669, An integrative framework for strategy making processes, #cademy of
2ana!ement <e,ie+, 57, pp;972;?5
Dart, S. and :anbury, #., 566<, Dow Strategy2+aking (rocesses #an +ake a
.ifferenceI, Strate!ic 2ana!ement Gournal, 5?, pp9?529@6
Dershey, ,.#., 4unreuther, D.#., and Schoemaker, (.,.D., 56=9, Sources of bias in
assessment procedures for utility functions, 2ana!ement Science, 9=, pp6;@26?<
Despos, $."., and Straussman, (.A., 56@?, Stochastic .ecision Trees in the Analysis of
Investment .ecisions, 2ana!ement Science, number 5>, pp9<<29?6
Digson, #., 566?, .usiness (inance, 9
nd
Adition, :utterworths, -ondon
Dirst, A., and Schweit1er, +., 566>, Alectric2utility resource planning and decision2
making, <isk #nalysis, 5>, pp5;725<@
Ditt, +.A. and Tyler, :.:., 5665, Strategic decision models: Integrating different
perspectives, Strate!ic 2ana!ement Gournal, @, pp57525=>
Dogarth, $.+., 56=7, )he psycholo!y of $ud!ement and choice, 9
nd
Adition, ,ossey2
:ass, San "rancisco
Dolmes, (., 566=, *n,estment #ppraisal, International Thomson :usiness (ress,
-ondon
DooverBs on2line business network, http:NNwww.hoovers.com
Dora, S.#., .odd, /.)., and Dora, ,.A., 566;, The use of decomposition in probability
assessments of continuous variables, Gournal of .eha,ioural :ecision-makin!, @,
pp5;;25<7
Dosseini, ,., 56=@, .ecision Analysis and its Application in the #hoice between Two
!ildcat %il 0entures, *nterfaces, 5@ &9', pp7?2=?
Doward, $.A., 56==, .ecision analysis: (ractice and (romise, 2ana!ement Science,
volume ;<, number @, ,une 56==, pp@762@6?
Doward, $.A., 56@=, )he foundations of decision analysis, IAAA Trans. Sys. Sc. #yb.
SS#2<, pp9552956
Dull, ,.#., 5677, .ealing with dependence in risk simulations, 1perational <esearch
Huarterly, 9= &5', pp9>5295;
99=
Dumphreys, (., 56=>, .ecision Aids: Aiding .ecisions in S8oberg, -., Tys1ka, T., !ise,
,.A., &editors', Duman :ecision-makin!, :odfors, .o*a
Dumphreys, (., and :erkeley, .., 56=?, Dandling uncertainty: -evels of analysis of
decision problems, in !right, )., &editor', .eha,ioural :ecision-makin!, (elnum
(ress, /ew Cork
Dumphreys, (.#., and +c"adden, !., 56=>, A*periences with +A3.: Aiding decision
structuring versus bootstrapping the decision2maker, #cta %sycholo!ica, <?, pp?52@6
Dyne, /., 566?, @one technical !uide to petroleum !eolo!y, e0ploration, drillin! and
production, (enwall (ublishing #ompany, Tulsa
Ikoku, #.3., 56=<, @atural !as reser,oir en!ineerin!, !iley, /ew Cork
,anis, I.-., 56=6, Crucial decisions: Leadership in policy makin! and crisis
mana!ement, "ree (ress, /ew Cork
,anis, I.-., and +ann, -., 5677, :ecision-makin!: a psycholo!ical analysis of conflict,
choice and commitment, "ree (ress, /ew Cork
,aynes, A.T., 56?@, %ro'a'ility theory +ith applications in Science and &n!ineerin!,
/o < in #olloquium -ectures in (ure and Applied Science, Socony2+obil %il
#ompany, 3SA
,effreys, :., 56;6, )heory of pro'a'ility, #laredon (ress, %*ford
,ohn, $.S., von !interfeldt, .., Adwards, !., 56=;, The quality and user acceptance of
multiattribute utility analysis performed by computer and analyst, in Dumphreys, (.,
von !interfeldt, .., Adwards, !., &editors, #nalysin! and aidin! decision processes,
pp;>52;56, Amsterdam.
,ones, $.A., ,acobs, -.!., and 0an Spi8ker, !., 5669, Strategic .ecision (rocesses in
International "irms, 2ana!ement *nternational <e,ie+, ;9, pp95629;@
4ahneman, .., Slovic, (. and Tversky, A., &editors', 56=9, Gud!ement =nder
=ncertainty: Deuristics and .iases, #ambridge 3niversity (ress
4ahneman, .., and Tversky, A., 56=9, The psychology of preferences, Scientific
#merican, 9<@, number 5, p5;@
4aufman, ).+., and Thomas, D., 5677, 2odern :ecision #nalysis, (enguin :ooks,
-ondon
4aye, .., 566?, The importance of information, 2ana!ement :ecision, volume ;;,
number ?, p?
4eats, :.!., 56==, The vertical construct validity of business economic performance
measures, )he $ournal of applied 'eha,ioural science, 9<&9', pp5?525@>
996
4eeffer, ..-., and :odily, S.A., 56=;, Three (oint Appro*imations for #ontinuous
$andom 0ariables, 2ana!ement Science, 96 &?', pp?6?2@>6
4eeney, $., 56=9, .ecision Analysis: an %verview, 1perations <esearch, ;>, number
?, pp=>;2=;7
4eeney, $., 56=>, .ecision Analysis in the )eo2technical and Anvironmental "ields, in
S8oberg, T., Tys1ka, T., and !ise, ,.A., &editors', Duman :ecision-makin!, :odafors,
.o*a
4eeney, $., 5676, .ecision Analysis: how to cope with increasing comple*ity,
2ana!ement <e,ie+, p9@
4eeney, $.-. and $aiffa, D., 567@, :ecisions +ith 2ultiple 1'$ecti,es: %references
and Calue )radeoffs, !iley, /ew Cork
4eeney, $.-. and $aiffa, D., 5679, A critique of formal analysis in public decision2
making, in .rake, A.!., 4eeney, $.-. and +orse, (., &editors', #nalysis of pu'lic
systems, pp6-55, +IT (ress, #ambridge
4emma, A.).P., 566;, #ase Studies on $eal %ptions, (inancial 2ana!ement, 99,
pp9?6297>
4eren, )., 566@, (erspectives of :ehavioural .ecision2making: Some critical notes,
1r!anisational .eha,iour and Duman :ecision %rocesses, @;&;', +arch, pp5@6257=
4ester, !.#., 56=<, TodayBs options for tomorrowBs growth, Dar,ard .usiness <e,ie+,
+arch J April
4eynes, ,.+, 5695, # treatise on pro'a'ility, +acmillan, -ondon
4im, S.D., "arragher, A.,. and #rick, T., 56=<, "oreign capital budgeting practices used
by the 3S and non23S multinational companies, &n!ineerin! &conomist, 96&;',
pp9>7295?
4imbrough, S.%., and !eber, +., 566<, An empirical comparison of utility assessment
programs, &uropean Gournal of 1perational <esearch, 7?, pp@572@;;
4leinmunt1, ../., 566>, .ecomposition and the control of error in decision analytic
models, in Dogarth, $. &editor', *nsi!hts in decision-makin!: # tri'ute to Dillel
G.&inhorn, pp5>7259@, 3niversity of #hicago (ress, #hicago
4leinmunt1, ../., "ennema, +.)., and (eecher, +.A., 566@, #onditional assessment
of sub8ective probabilities: Identifying the benefits of decomposition, 1r!anisational
.eha,iour and Duman :ecision %rocesses, @@, pp525?
4rant1, .., -uce, $..., Suppes, (. and Tversky, A., 5675, (oundations of
measurement, 0olume 5, Academic (ress, /ew Cork
9;>
4unreuther, D.#., and Schoemaker, (.,.D., 56=>, :ecision analysis for comple0
systems inte!ratin! descripti,e and prescripti,e components, !orking paper,
.epartment of decision sciences, 3niversity of (ennsylvania
-aherrbre, ,., D., 5666, $eserve )rowth: Technological (rogress or :ad $eporting and
:ad Arithmetic, -eopolitics of &ner!y, April
-aing, A., 5667, 2arketin! in @DS )rusts: #doption and #daptation of 2arketin!
Concepts in a %u'lic Sector Settin!, unpublished (h... Thesis, 3niversity of Aberdeen
-amb, ".A., Simpson, ).S., "inch, ,.D., 5666, +ethods for evaluating the worth of oil
and gas reserves, -eopolitics of &ner!y, April
-angley, A., 566?, :etween Hparalysis by analysisK and He*tinction by instinctK, Sloan
2ana!ement <e,ie+, volume ;?, number ;, pp@;277
-angley, A., 56=6, In search of rationality: The purposes behind the use of formal
analysis in organisations, #dministrati,e Science Huarterly, ;<, pp?6=2@;5
-aplace, (.S., 5=59, %hilosophical essay on pro'a'ility, (aris, #hapter 5=
-athrop, ,.!., and !atson, S.$., 56=9, .ecision analysis for the evaluation of risk in
nuclear waste management, Gournal of 1perational <esearch Society, ;;, pp<>72<5=
-each, #., 5676, *ntroduction to Statistics, !iley, Surrey
-efley, "., and +organ, +., 5666, The /(0 (rofile: a creative way of looking at /(0,
2ana!ement #ccountin!, ,une, volume 77, issue @, p;6
-ehmann, A.-., and .BAbrera, D.,.+., 567?, @on-parametrics: Statistical 2ethods
.ased on <anks, Dolden2.ay Inc., San "rancisco
-eonard2:arton, .., 566>, A .ual +ethodology for #ase Studies: Synergistic use of a
-ongitudinal Single Site with $eplicated +ultiple Sites, 1r!anisational Science,
volume 5, number ;, pp9<=29@@
-eslie, 4.,., and +ichaels, +.(., 5667, The $eal (ower of $eal %ptions, 2c>insey
Huarterly, number ;
-evin, ,. and "o*, ,., 56==, &lementary statistics in social research, fourth edition,
Darper and $ow,
-ichenstein, S., "ischoff, :., and (hillips, -..., 56=9, #alibration of (robabilities: The
State of the Art to 56=>, in 4ahneman, .., Slovic, (., and Tversky, A., &editors',
Gud!ement =nder =ncertainty: Deuristics and .iases, #ambridge 3niversity (ress,
/ew Cork
-ichenstein, S., and Slovic, (., 5675, $eversals of preference between bids and choices
in gambling decisions, Gournal of &0perimental %scyholo!y, =6, pp <@2??
9;5
-indman, D.$., 5675, Inconsistent preferences among gambles, Gournal of
&0perimental %sycholo!y, =6, pp;6>2;67
-ipshit1, $., and Strauss, %., 5667, #oping with uncertainty, 1r!anisational .eha,iour
and Duman :ecision %rocesses, @6 &9', pp5<625@;
-ipshit1, $., and :en Shaul, %., 5667, Schemata and mental models in recognition
primed decision2making, in Psambok, #., 4lein, ).A, &editors', @aturalistic :ecision-
makin!, pp9692;>;, Dillsdale, /ew ,ersey, Arlbaum
-ohren1, ,., and .ickens, $./., 566;, %ption Theory for Avaluation of %il and )as
Assets: The 3psides and .ownsides, S%& 54A67, paper prepared for presentation at the
Society of (etroleum Angineers Dydrocarbon and Avaluation Symposium held in
.allas, Te*as, 962;> +arch
-othian, /., 56=7, 2easurin! corporate performance, The #hartered Institute of
+anagement Accountants
+ac#rimmon, 4.$., and !ehrung, ..A., 56=@, )akin! <isks, "ree (ress, /ew Cork
+ac)regor, .., and -ichenstein, S., 5665, (roblem structuring aids for quantitative
estimation, Gournal of 'eha,ioural decision-makin!, <, pp5>5255@
+ac)regor, .., -ichenstein, S., and Slovic, (., 56==, Structuring knowledge retrieval:
An analysis of decomposing quantitative 8udgements, 1r!anisational .eha,iour and
Duman :ecision %rocesses, <9, pp;>;2;9;
+adden, T.,., Dyrnick, +.#., Dodde, ,.A, 56=;, .ecision Analysis use to evaluate air
quality control equipment for the %hio Adison %il company, *nterfaces, 5;&5', pp@@27?
+arch, ,.)., and Simon, D.A., 56?=, 1r!anisations, !iley, /ew Cork
2ana!ement #ccountin!, 5667, ".s call for new criteria for measuring company
performance, September, volume 7?, number =, p5>
+arkides, #.#., 5666, A dynamic view of strategy, Sloan 2ana!ement <e,ie+, Spring,
volume <>, issue ;, p??
+arkland, ,.T., 5669, %ptions Theory: A new way forward for e*ploration and
engineering economicsI, S%& 5B565, paper prepared for presentation at the Society of
(etroleum Angineers %il ^ )as Aconomies, "inance and +anagement #onference
held in -ondon, 9@296
th
April
+arkowit1, D. +., 56?9, %ortfolio Selection, !iley, /ew Cork
+ason S., and +erton, $.#., 56=?, The role of contingent claims analysis in corporate
finance, in Altman, A. and Subrahmanyam &editors', <ecent #d,ances in Corporate
(inance, Domewood, Il: $ichard .. Irwin
9;9
+aykut, (., and +orehouse, $., 566<, .e!innin! Hualitati,e <esearch # philosophical
and practical !uide, The "almer (ress
+c#askey, +.:., 56=@, )he e0ecuti,e challen!e: 2ana!in! chan!e and am'i!uity,
Darper :usiness (ress
+c#unn, (., 566=. The balanced scorecard, 2ana!ement #ccountin!, volume 7?, issue
55, p<>
+eehl, (.A., 567=, Theoretical risks and tabular asterisks, Sir 4arl, Sir $onald, and the
slow progress of soft psychology, Gournal of Counsellin! and Clinical %sycholo!y, <@,
pp=>@2=;<
+egill, $.$., 5676, #n introduction to e0ploration economics, Second edition, Tulsa,
(etroleum (ublishing #ompany, p5=>
+egill, $.A., 5675, #n introduction to e0ploration economics, "irst edition, Tulsa,
(etroleum (ublishing #ompany, p5?6
+iles, $.D., 56=9, Coffin nails and corporate strate!ies, Anglewood #liffs, /ew
,ersey, (rentice Dall
+iles, +.:., 5676, Gualitative .ata as an attractive nuisance, #dministrati,e Science
Huarterly, 9<, pp?6>2@>5
+illiken, ".#., 56=7, Three types of perceived uncertainty about the environment:
State, effect, and response uncertainty, #cademy of 2ana!ement <e,ie+, 59&5', pp5;;2
5<;
+ilne, ,.,. and #han, #.#.#., 5666, /arrative corporate disclosure: how much of a
difference do they make to investment decision makingI, )he .ritish #ccountin!
<e,ie+, volume ;5, number <, pp<;62<?7
+int1berg, D., $aisinghani .., and Theoret A., 567@, The Structure of HunstructuredK
decision processes, #dministrati,e Science Huarterly, 95, ,une, p9<@297?
+oody2Stuart +., 5666, %rosperin! under uncertainty / the @orth Sea in e,ol,in!
markets, speech given at %ffshore Aurope, September 7th
.e +organ, A., 5=<7, (ormal lo!ic - )he calculus of inference: necessary and
pro'a'le, Taylor ^ !alton, -ondon
+organ, +.). and Denrion, +., 566>, =ncertainty: # -uide to :ealin! +ith
=ncertainty in Huantitati,e <isk and %olicy #nalysis, #ambridge 3niversity (ress
+owen, ,.#., and )entry, ,.!., 56=>, Investigation of the preference reversal
phenomenon in a new product introduction task, Gournal of #pplied %sycholo!y, @?,
pp75?2799
9;;
+urtha, ,.A., 5667, +onte #arlo simulation: its status and future, S(A ;76;9, Gournal
of %etroleum )echnolo!y, <6, p;@5.
+yers, S.#., 56=<, "inancial Theory and "inancial Strategy, *nterfaces, number 5<,
,anuary 2 "ebruary, pp59@25;7
+yers, S.#, 5677, The .eterminants of #orporate :orrowing, Gournal of (inancial
&conomics, ?, /ovember
+yers, ..)., and -amm, D., 567?, The polarising effect of group discussion,
#merican Scientist, @;, pp9672;6;
/angea, A.)., and Dunt, A.,., 5667, An Integrated .eterministicN(robabilistic
Approach to $eservoir Astimation: An 3pdate, S%& 6AA96, presented at S(A Annual
Technical #onference and A*hibition, San Antonio, Te*as, ?2= %ctober
/ewendorp, (..., 566@, :ecision #nalysis for %etroleum &conomics, second edition,
(lanning (ress, Aurora #olorado
/isbett, $., and $oss, -., 56=>, Duman *nference: Strate!ies and shortcomin!s of
social $ud!ement, (rentice Dall, Anglewood #liffs, /ew ,ersey
/utt, (.#., 5666, Surprising but true: Dalf of the decisions in organisations fail, )he
#cademy of 2ana!ement &0ecuti,e, /ovember, volume 5;, issue <, p7?
/utt, (.#., 5667, :etter decision2making: a field study, .usiness Strate!y <e,ie+,
!inter, volume =, number <, p<?
/utt, (.#., 566;, The formulation processes and tactics used in organisational decision2
making, 1r!anisation Science, <, pp99@29?5
%blak, ..,. and Delm, $.,., ,r., 56=>, Survey and analysis of capital budgeting
methods used by multinationals, (inancial 2ana!ement, !inter, pp;72<5
%B+ahoney, ,., 4., 566=, 2essy 2ethodolo!ies: # Case Study, !orking (aper ;>;,
!arwick :usiness School, 3niversity of !arwick
%dell, (., 566?, The world is running out of oil not into it, *nternational &ner!y
3orkshop held at the *nstitute for applied system analysis, 9>299 ,une
%kley, ,., 566<, Thinking through fieldwork, in :ryman, A. and :urgess, $.,).,
editors, #nalysin! Hualitati,e :ata, pp5=2;;, $outledge, -ondon
%ransanu, ,., and #onnolly, T., 566;, The reinvention of decision2making, in 4lein,
).A., %ransanu, ,., #alderwood, $., and Psambok, #., &editors', :ecision-makin! in
action: 2odels and methods, pp;29>, Able* (ublishing, /orwood, /ew ,ersey
9;<
%rdone1, -., and :enson, -., 5667, .ecisions under time pressure: how time
constraints affect risky decision2making, 1r!anisational .eha,iour and Duman
:ecision %rocesses, 75 &9', August, pp59525<>
%rton, ,., 5667, "rom inductive to iterative, Scandina,ian Gournal of 2ana!ement,
volume 5;, number <, pp<562<;=
%tely, .., 5667, :etter (erformance +anagement, 2ana!ement #ccountin!, volume
7<, number 5, p<<
%tis, $.+., and Schneidermann, /., 5667, A (rocess for Avaluating A*ploration
(rospects, )he #merican #ssociation of %etroleum -eolo!ists .ulletin, volume =5,
number 7, ,uly, pp5>=7255>6
(addock, ,. and Siegel, ..$., and Smith, ,.-., 56==, %ption 0aluation of claims on real
assets: The case of offshore petroleum leases, )he Huarterly Gournal of &conomics,
August, pp<762?>7
(apadakis, 0., 566=, Strategic Investment .ecision (rocesses and %rganisational
(erformance: An Ampirical A*amination, .ritish Gournal of 2ana!ement, volume 6,
number 9, pp55?25;;
(apadakis, 0., and -ioukas, S., 566@, .o early perceptions of strategic decisions
influence strategic processesI An empirical investigation, #cademy of 2ana!ement
.est %aper %roceedin!s, #incinnati, %hio, pp<@2?>
(earce, ,.A. II, $obbins, 4..., and $obinson, $.:. &,r.', 56=7, The Impact of )rand
Strategy and (lanning "ormality on "inancial (erformance, Strate!ic 2ana!ement
Gournal, =, pp59?25;<
(earson, A.S., and Tukey, ,.!., 56@?, Appro*imating +eans and Standard .eviations
:ased on .istances between (ercentage (oints of "requency #urves, .imetrika, ?9,
number ;2<, pp?;;2?<@
(eterson, #.$., and :each, -.$., 56@7, +an as an intuitive statistician, %sycholo!ical
.ulletin, @=, pp962<@
(ettigrew, A., 567;, )he politics of or!anisational decision-makin!, -ondon, Tavistock
(feffer, ,., 56=5, %o+er in or!anisations, +arshfield, +A: (itman
(feffer, ,., and Salancik, ).$., 567=, )he e0ternal control of or!anisations: # resource
dependence perspecti,e, Darper and $ow, /ew Cork
(hillips, -..., 56=6, .ecision analysis in the 566>Bs, in Shahini, A. and Stainton, $.,
&editors', )utorial papers in operational research 19A9, %perational $esearch Society
9;?
(ike, $. and /eale, :., 5667, Corporate (inance and *n,estment. :ecisions and
Strate!ies, 9
nd
Adition, (rentice Dall, -ondon.
(olyani, +., 56@@, %ersonal >no+led!e: )o+ard a %ost Critical %hilosophy,
3niversity of #hicago (ress, #hicago
(riem, $.-., 566>, Top +anagement Team )roup "actors, #onsensus, and "irm
(erformance, Strate!ic 2ana!ement Gournal, 55, pp<@62<7=
(riem, $.-., $asheed, A.+.A., and 4otulic, A.)., 566?, $ationality in Strategic
.ecision2making (rocesses, Anvironmental .ynamism and "irm (erformance,
Strate!ic 2ana!ement Gournal, 55, pp<@62<7=
(rudential Securities, 9>>>, http:NNwww.prudentialsecurities.comN
Guinn, ,.:., 56=>, Strate!ies for chan!e: Lo!ical incrementalism, Domewood, Il:
Irwin
$aiffa, D., 56@=, :ecision analysis: *ntroductory Lectures on choices under
uncertainty, Addison2!esley
$aiffa, D., and Schlaifer, $., 56@5, #pplied statistical decision theory, Darvard
:usiness School (ress, +assachusetts
$a8agopalan, /., $asheed, A.+.A., .atta, ..4., 566;, Strategic .ecision2making
(rocesses: #ritical $eview and "uture .irections, Gournal of 2ana!ement, volume 56,
number 9, pp;<62;=<
$amsey, ".(., 569@, Truth and probability, reprinted 4yburg, D.A., and Smokler, D.A.,
&editors', Studies in su'$ecti,e pro'a'ility, 56@;, !iley
$ose (.$., 56=7. .ealing with risk and uncertainty in e*ploration: how can we
improveI, )he #merican #ssociation of %etroleum -eolo!ists .ulletin, volume 75,
number5, pp525@.
$oss, .., 5667, %ortfolio mana!ement applied in the oil industry, T/% internal report,
/IT) 67296;2:
$ossman, ). :., and $allis, S."., 566=, Learnin! in the field: #n introduction to
"ualitati,e research, Sage, Thousand %aks, #alifornia
Saaty, T.-., 56=>, )he #nalytic Dierarchy %rocess, %lannin! %riority Settin! and
<esource #llocation, +c2)raw Dill, /ew Cork
Savage, -.,., 56?<, )he foundations of statistics, !iley, /ew Cork
Schafer, $.A., and :orcherding, 4., 567;, The assessment of sub8ective probability
distributions: A training e*periment, #cta %sycholo!ica, ;7, pp5572596
9;@
Schendel, ..A., and Dofer, #.!., editors, 5676, Strate!ic 2ana!ement: a @e+ Cie+ of
.usiness %olicy and %lannin!, -ittle :rown, :oston
Schoemaker, (.,.D., 566?, Scenario (lanning: A tool for strategic thinking, Sloan
2ana!ement <e,ie+, ;@&9', pp9?2?>
Schuyler, ,.$., 5667, :est practices in pro8ect evaluation and influence on company
performance, Gournal of %etroleum )echnolo!y, August, pp=5=2=;<
Schweiger, ..+., Sandberg, !.$., and $agan, ,.!., 56=@, )roup Approaches for
Improving Strategic .ecision2making: A comparative analysis of .ialectical Inquiry,
.evilBs Advocacy, and #onsensus, #cademy of 2ana!ement Gournal, 96&5', pp?7275
Seaver, .. A., 567=, Assessing (robabilities with +ultiple Individuals: )roup
Interaction 0ersus +athematical Aggregation, )echnical <eport SS<*-7A-6, Social
Science $esearch Institute, 3niversity of South #alifornia, -os Angeles
Shackle, ).-.S., 567<, .ecision: The human predicament, #nnals of the #merican
#cademy of %olitical and Social Science, <59, +arch
Shaffir, !., :., Stepbins, $., A., and Turowet1, A., 56=>, editor, (ield+ork
&0perience: Hualitati,e #pproaches to Social <esearch, St. +artinBs (ress, /ew Cork
Shao, -.(., and Shao, A.T., 566;, #apital budgeting practices employed by Auropean
affiliates of 3S transnational companies, Gournal of 2ultinational (inancial
2ana!ement, ; &5N9', pp6?25>6
Shell, 566=, Shell .rent, 7
th
%ctober, speech given at the Shell -eadership conference
Shipman, +., 567@, &editor', )he 1r!anisation and *mpact of Social <esearch,
$outledge and 4egan (aul, -ondon
Siegel, ..$. , Smith, ,.-., and (addock, ,.-., 56=7, 0aluing %ffshore %il (roperties
with %ptions (ricing +odels, 2idland Corporate (inance Gournal 4, pp992;>
Siegel, S. and #astellan, /.,., 56==, @onparametric statistics for the 'eha,ioural
sciences, 9
nd
edition, +c)raw2Dill, /ew Cork
Simpson, ).S., -amb, ".A., "inch, ,.D., .innie, /.#., 9>>>, The application of
probabilistic and qualitative methods to asset management decision2making, S%&
49B44, prepared for S(A Asia (acific #onference on Integrated +odelling for Asset
+anagement Cokohama 9?29=
th
April
Simpson, ).S., "inch, ,.D. and -amb, ".A., 5666, $isk, uncertainty and relations
between HstrategicK and Hwithin strategyK decision2making in the upstream oil and gas
industry, in %roceedin!s of the B
th
international conference on the dynamics of strate!y,
3niversity of Surrey, pp;;?2;?=.
9;7
Singh, ). and 4inagi, )., 56=7, <isk and relia'ility appraisal on 2icro Computers,
#hartwell2:rat, -und
Slovic, 4.)., 566?, The construction of preference, #merican %sycholo!ist, ?>, pp;@<2
;75
Smith, 4.)., )annon, +.,., )rimm, #. and +itchell, T.$., 56==, .ecision2making
behaviour in smaller entrepreneurial and larger professionally managed firms, Gournal
of .usiness Centurin!, ;, pp99;29;9
Smith, ,.A., and +c#ardle, 4., 5667, %ptions in the real world: -essons learned in oil
and gas investments, "uquaN.uke 3niversity, !orking paper, April
Smithson, +., 56=6, *!norance and uncertainty: &mer!in! paradi!ms, Springer
0erlag, /ew Cork
Snow ,.D., .orb A.)., and .orn2-ope1 ..!., 566@, $isk analysis and full2cycle
probabilistic modelling of prospects: A prototype system developed for the /orwegian
Shelf, in .orb A.). and Sinding2-arsen $., eds., Huantification and %rediction of
%etroleum <esources, /(" special publication number @5, Alsevier, Amsterdam,
pp5?;25@?
Society of (etroleum Angineers, !orld (etroleum #ongress and American Association
of (etroleum )eologists, 9>>>, http:NNwww.spe.orgNcdaNcontentN>,5<9=,9=?,>>.htm
Spradlin, T., 5667, # le0icon of decision-makin!,
http:NNwww.faculty.fuqua.duke.edu NdawebNle*icon.htm
Spencer, ,.A., and +organ, ..T.4., 566=, Application of "orecasting and 3ncertainty
+ethods to (roduction, S%& B9995, presented at the S(A Annual Technical #onference
and A*hibition, /ew %rleans, -ouisiana, 972;> September
Stael val Dolstein, #.A.S, 5675a, The effect of learning on the assessment of sub8ective
probability distributions, 1r!anisational .eha,iour and Duman :ecision %rocesses, @,
pp;><2;5?
Stael val Dolstein, #.A.S, 5675b, Two techniques for assessment of sub8ective
probability distributions: An e*perimental study, #cta %sycholo!ica, ;?, pp<7=2<6<
Stanley, +.T., and :lock, S.:., 56=;, An empirical study of management and financial
variables influencing capital budgeting decisions for multi2national companies in the
56=>s, 2ana!ement *nternational <e,ie+, 9; &;', pp@5279
Stonehill, A.I., and /athanson, -., 56@=, #apital budgeting and the multinational
corporation, California 2ana!ement <e,ie+, Summer, pp;62?<
9;=
Swalm, $.%., 56@@, 3tility Theory J Insights into $isk Taking, Dar,ard .usiness
<e,ie+, /ovember2.ecember, pp59;25;@
The Aconomist &3S', 566=, 1il industry: )he decade7s +orst stocks, volume ;<6, issue
=>67, pp7;27<
The Aconomist, 566@, (rom ma$or to minor, volume ;;6, number 76@@, +ay 5=
th
,
pp@;2@?
Thomas, D., and Samson, .., 56=@, Sub8ective aspects of the art of decision analysis:
e*ploring the role of decision analysis in decision structuring, decision support and
policy dialogue, Gournal of 1perational <esearch Society, ;7, pp9<629@?
Thompson, ,., 56@7, 1r!anisations in action, +c)raw Dill, /ew Cork
Tocher, 4..., 567=, 0iewpoints, Gournal of 1perational <esearch Society, 96, pp5762
5=9
Tocher, 4..., 5677, (lanning Systems, %hilosophical )ransactions of the <oyal
Society of London, A9=7, pp<9?2<<5
Tocher, 4..., 567@, /otes for discussion on HcontrolK, 1perational <esearch
Huarterly, 97 pp9;529;6
Trigeorgis, -, 566@, <eal 1ptions: 2ana!erial (le0i'ility and Strate!y in <esource
#llocation, +assachusetts Inst. of Technology (ress, #ambridge, +assachusetts
Tversky, A. and 4ahneman, .., 56=5, The framing of decisions and the psychology of
choice, Science, 955, pp<?;2<?=
Tversky, A. and 4ahneman, .., 567<, ,udgement 3nder 3ncertainty: Deuristics and
:iases, Science, 5=?, pp559<255;5
3liva, ,.!., 56=7, (ost Automation &PipU<' Technology: A .ecision Analysis,
*nterfaces, 79 &9', pp5259
3nited 4ingdom %ffshore %perators Association, 9>>>a, )he 1ri!ins of 1il and -as,
http:NNwww.ukooa.co.uk
3nited 4ingdom %ffshore %perators Association, 9>>>b, Current <eser,es and
%ro$ected Life, http:NNNwww.ukooa.co.ukNpresentNharness9.html
3nited 4ingdom %ffshore %perators Association, 9>>>c &conomic Contri'ution 1999,
http:NNwww.ukooa.co.uk
0an de 0en, A., D., 5669, Suggestions for Studying Strategy (rocess: A research note,
Strate!ic 2ana!ement Gournal, volume 5;, pp5@625==
0an de 0liet, A., 5667, The new balancing act, 2ana!ement )oday, ,uly, p7=
9;6
0enkatraman, /., and $amanu8am, 0., 56=7, +easurement of :usiness Aconomic
(erformance: an e*amination of +ethod #onvergence, Gournal of 2ana!ement, 5;&5',
pp5>62599
0on /eumann, ,., and +orgenstern, %., 56<7, )heory of !ames and economic
'eha,iour, 5
st
edition, (rinceton
0on /eumann, ,., and +orgenstern, %., 56<7, )heory of !ames and economic
'eha,iour, 9
nd
edition, (rinceton
0on !interfeldt, ..0., 56=>, Structuring decision (roblems for .ecision Analysis,
#cta %sycholo!ica, <?, pp7;26;
!ald, A., 56@<, Statistical decision functions, !iley, /ew Cork
!atson (., 566=, A process for estimating geological risk of petroleum e*ploration
prospects, #ustralian %etroleum %roduction and &0ploration #ssociation Gournal,
p?77 J ?=;.
!eick, 4.A., 5676, Sensemakin! in or!anisations, Sage, Thousand %aks, #A
!ells, ).A., 56=9, The use of decision analysis in Imperial )roup, ,ournal of
%perational $esearch Society, ;;, pp;5;2;5=
!ensley, $., 5666, (roduct strategies, managerial comprehension and organisation
performance, 10ford <e,ie+ of &conomic %olicy, Spring, volume 5?, issue 5, p;;
!ensley, $., 5667, A*plaining success: the rule of ten percent and the e*ample of
market share, .usiness Strate!y <e,ie+, Spring, =&5', p@;
!eston, ,."., and :righman, A."., 567=, 2ana!ement (inance, @
th
edition, Dolt,
$inehart and !inston, /ew Cork
!hiteside, +., 5667, (ortfolio %ptimisation, *nde,a &ner!y Consultants, %*on
!hitney, ,. and Smith, $., 56=;, Affects of #ohesiveness on Attitude (olarisation and
the Acquisition of 4nowledge in a Strategy (lanning #onte*t, Gournal of 2arketin!
<esearch, 9>, pp5@7257@
!hyte, !."., 56??, Street Corner Society, 9
nd
Adition, 3niversity of #hicago (ress,
#hicago
!icks 4elly, +.A. and (hilippatos, ).#., 56=9, #omparative analysis of foreign
investment practices by 3S based manufacturing multinational companies, Gournal of
*nternational .usiness Studies, 5;&;', pp562<9
!ildavski, A., 56==, Searchin! for safety, Transactions (ublishers, /ew :runswick
&3SA'
9<>
!ilmot, (., .ewynne, ,., and Dowison, S., 566<, 1ption %ricin!, "inancial (ress,
%*ford
!inter, ".!., 56=?, An Application of #omputerised .ecision Tree +odels in
+anagement 3nion :argaining, *nterfaces, 5?&9', pp7<2=>
!iseman, ,., 567<, The research web, =r'an Life and Culture, volume ;, number ;,
pp;572;9=
!ohstetter, $., 56@9, %earl Dar'our: 3arnin! and :ecision, Stanford 3niversity
(ress, Stanford, #A
!ood +acken1ie, 9>>>, @e+sletter, issue ;7, "ebruary
!ood +acken1ie, 5666, @orth 3est &urope =pstream Ser,ice, +ay
!ood +acken1ie, 566=, @orth Sea <eport 575, /ovember
!ooldridge, :., and "loyd, S.!., 566>, The Strategy (rocess, +iddle +anagement
Involvement, and %rganisational (erformance, Strate!ic 2ana!ement Gournal, 55,
pp9;529<5
!orld 0est :ase, http:NNwww.wvb.com
!right Investors Service, http:NNwww.wisi.comNinde*.htm
Cates, ,.". and Stone, A.$., 5669, The risk construct, in Cates, ,."., &editor', <isk takin!
'eha,iour, pp529@, !iley, /ew Cork
Cergin .., 5665, )he %rize: )he epic "uest for oil, money and po+er, Simon and
Schuster, Sydney
Pimmer, A.#., 56=;, 0erbal vs /umerical processing of sub8ective probabilities, in
Schol1, $.!., &editor', :ecision-makin! under uncertainty, Amsterdam
9<5