You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/290821330

Do personality traits affect decision making ability: Can MBTI type predict
biases?

Article  in  The APPEA Journal · January 2011


DOI: 10.1071/AJ10022

CITATIONS READS

5 4,664

5 authors, including:

Matthew Brian Welsh S. H. Begg


University of Adelaide University of Adelaide
68 PUBLICATIONS   622 CITATIONS    94 PUBLICATIONS   998 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Matthew Brian Welsh on 18 January 2016.

The user has requested enhancement of the downloaded file.


Lead author
DO PERSONALITY TRAITS AFFECT
Matthew DECISION MAKING ABILITY:
Welsh
CAN MBTI TYPE PREDICT BIASES?

M. Welsh, A. Alhakim, F. Ball, J. Dunstan and S. Begg by ensuring that the right people are employed to make
University of Adelaide decisions where bias may prove to be a problem.
North Terrace To test this, we distributed a survey—which included
Adelaide SA 5005 the complete Myers-Briggs test and questions to identify
biases in the respondent’s responses—to engineering stu-
matthew.welsh@adelaide.edu.au
dents at the University of Adelaide, petroleum industry
abdulaziz.alhakim@student.adelaide.edu.au
employees and a small number of employees from other
finlay.ball@student.adelaide.edu.au industries. Individual MBTI distinctions (e.g., extroverts
joseph.dunstan@student.adelaide.edu.u versus introverts) were studied for bias tendencies. We
steve.begg@adelaide.edu.au discuss the observed relationships between Myers-Briggs
type and decision biases and their relevance to decision
ABSTRACT making in the petroleum industry. We conclude that
while there do seem to be some effects of personality on
From decades of psychological research, the observa- susceptibility to decision-making biases, these are not
tion that people’s decisions are often biased by particular necessarily what one might expect given the descriptions
decision-making flaws has led to discussions of what can attached to specific MBTI types. They are also relatively
be done to de-bias decisions. A key area of research is the weak effects and, as such, the use of the Myers-Briggs as a
study of individual differences in decision-making abil- tool for assessing decision-making ability is not supported
ity—that is, whether certain people are less susceptible by our results.
to particular biases. Much of this research, however, has
focused on the impact of intelligence on decision-making KEYWORDS
ability. This, however, is of limited use in industries such
as oil and gas where, due to hiring criteria that commonly Decision making, overconfidence, framing, personality,
include at least a bachelor’s degree, a restricted range of biases, Myers-Briggs.
intelligence is observed. Given this, it may be more fruit-
ful to consider other sources of potential differences in INTRODUCTION
decision-making ability such as personality traits.
The Myers-Briggs Type Indicator (MBTI) (Myers et al, The ability to make accurate, unbiased decisions is a
1998) is a personality test based on Jung’s 1921 theory of highly valued attribute in industry contexts. Forty years of
psychological types (see, e.g., Jung, 1971). People are sorted psychological research, however, has demonstrated that this
into one of 16 categories based on their responses to the ability is far rarer than we might hope. Instead, research-
personality test. Although the test is widely used to identify ers—starting with Tversky and Kahneman’s (1974) seminal
leadership styles and preferences and therefore influence work—have repeatedly demonstrated that people’s judg-
recruitment decisions (including within the petroleum ments and decisions tend to be influenced by the heuristics
industry), the impact of Myers-Briggs personality type (simple decision rules) that we rely upon, which result
on decision making itself and, in particular, on decision- in systematic biases in the decision that we make. These
making biases, has not been thoroughly investigated. biases affect experts as well as novice decision makers
A large number of specific, cognitive biases have been (Northcraft and Neale, 1987; Welsh et al, 2005) and, when
identified by psychologists and in this study we have cho- affecting estimates of parameters used to feed into models
sen to investigate two of particular interest in the oil and for calculating reservoir size and development needs, can
gas industry: overconfidence and risk attitude/framing. It have significant impacts on the net present value (NPV)
has been shown (Welsh and Begg, 2008; Welsh et al, 2007) of oil and gas projects (Welsh, et al., 2007).
that these biases can have disastrous consequences on the Given this, significant work has looked at methods for
value of projects in the petroleum industry. If it is possible, de-biasing decisions: that is, methods for reducing the im-
therefore, to use an easily administered and well-known pact that specific biases have on estimates and decisions.
test (MBTI) to quantify and predict an individual’s suscep- Many of these include changes to the methods being used
tibility to bias, we can improve decisions in the industry to elicit estimates from experts. For example, there is a

SECOND PROOF—WELSH 10 FEB 11 APPEA Journal 2011—1


M. Welsh, A. Alhakim, F. Ball, J. Dunstan and S. Begg

sizeable literature devoted to ongoing argument about the world and make decisions. It yields scores on four
the best methods for asking for range (see, e.g., Block and scales: extraversion-introversion (E/I); sensing-intuition
Harper, 1991; Juslin et al, 1999; Russo and Schoemaker, (S/N); thinking-feeling (T/F); and judging-perceiving (J/P).
1992; Welsh et al, 2004). Standard practice is then to categorise a person based on
In general, however, while these de-biasing techniques do which end of each scale they fall closer to. So, for example,
provide a benefit, they tend to not eliminate bias, instead a person who scores closer to the extraversion end of the
only reduce it (Morgan and Henrion, 1990). As a result, E/I scale is labelled an E. In this way a person is sorted
other researchers have taken a different approach to the into one of the 16 possible four-letter types—an ESTJ,
problem. Specifically, rather than looking for methods that for example.
reduce bias in everyone, they look for people who are the Given the nature of the personality preferences detailed
least susceptible to bias in the first place. The key observa- in the MBTI, one would expect that people with certain
tion that these researchers rely on is that susceptibility to preferences might have different approaches to solving
bias in decision making is not evenly distributed among the decision-making problems and, as a result, have differential
population—instead, a common observation is that while susceptibilities to certain cognitive biases. In particular, the
most people are susceptible to any chosen bias, there are MBTI suggests that the perceiving sensory-intuitive (S/N)
always a few people who seem not to be. This leads to the dimension affects how people prefer to gather informa-
question: is it the same people in each case who are least tion, while the judging thinking-feeling (T/F) dichotomy
susceptible and, if so, can we work out why? indicates how people prefer to make their decisions.These
Some progress towards answering this question has al- distinctions bear some resemblance to the work of Stanov-
ready been made, with an extended research program by ich and West (2008) on two-system approaches to decision
Stanovich and West (1998, 2008) demonstrating that more making, where people are thought to make decisions us-
intelligent people are, to a small degree, less susceptible ing either an intuitive system or a rational system. Given
to certain biases. This finding is, however, of limited impor- this, it seem reasonable to expect that people in the N
tance to industries such as oil and gas looking to improve and F categories may be more susceptible to biases as
the decision-making ability of their personnel, primarily these people should prefer to make their decisions using
because most of the decision makers hired by oil compa- System 1 (simple, intuitive) processes.
nies are already sourced from among highly intelligent
people—university graduates. Given the relatively small Cognitive biases
benefits already described by Stanovich and West, this
truncation of the range of intelligence means that the dif- A wide range of specific cognitive biases have been
ferences in decision-making ability between the least and demonstrated to affect decisions both generally and spe-
most bright of industry personnel that can be accounted cifically in oil and gas contexts (see, e.g., Tversky and
for by intelligence are likely to be very small indeed. Kahneman, 1974; Welsh et al, 2005). A central question,
As a result, if our goal is to improve industry decision however, is: which of these are most likely to be affected
making through a greater understanding of individual by personality traits? A brief overview of the various biases
differences in decision-making ability, we need to look highlights two key suspects: overconfidence, and attitudes
at characteristics beyond intelligence. An obvious candi- to risk and uncertainty.
date is personality. While personality traits undoubtedly Overconfidence (Lichtenstein et al, 1982) in decision
play a role in determining a person’s career path, they do making-literature most commonly refers to the observed
not necessarily act as a bottleneck or hurdle in the same tendency of people to give ranges that contain the true
fashion as intelligence for industry employment. That is, value less often than they believe they will. This is of
because personality traits such as extraversion or introver- particular importance to the oil and gas industry where
sion are independent of intelligence, both introverts and calculations of oil in place are based on the estimated
extraverts are capable of meeting the educational criteria P10–P90 ranges provided by technical specialists for each
for employment within the industry. geological parameter. These ranges have been shown to
contain the true value (once it becomes known) less than
Myers-Briggs Type Indicator 50% of the time, rather than the expected 80% (Hawkins
et al, 2002). Confidence (and therefore overconfidence)
The Myers-Briggs Type Indicator (Myers et al, 1998) is seems a likely candidate to be affected by personality
a widely used personality test based on Jung’s theories of traits. In particular, one might expect that extraverted
psychological type (Jung, 1971).While not considered state- people (Es) might be more confident than introverts (Is).
of-the-art by psychologists (see, e.g., McCrae and Costa, A willingness to accept some risk is necessary in all
1989), it has nevertheless remained extremely popular as corporate ventures and the oil industry is, in some senses,
a recruitment and assessment tool in industries including riskier than most. However, biases in risk attitude—that
the oil and gas industry and was therefore selected for is, in technical terms, being risk seeking or risk averse
use here. rather than risk neutral—always erode value in the long
The MBTI uses a 94 item, forced-choice questionnaire term and therefore are an important consideration when
to measure a person’s preferences for how they perceive talking about improving decision making. These attitudes

2—APPEA Journal 2011 SECOND PROOF—WELSH 10 FEB 11


Do personality traits affect decision making ability: can MBTI type predict biases?

to risk and uncertainty also seem likely to be related to them to be categorised as one of 16 personality types.
personality rather than intelligence and, once again, one Section 2 of the survey consisted of questions assessing
might expect extraverts to be more risk seeking. Previous various biases, as described below.
work on this question using the MBTI by Li and Liu (2008) Section 2 included 10 range estimation questions, asking
supported this conjecture, also finding that more intui- for an 80% confidence range for a value. That is, partici-
tive and more feeling people (Ns and Fs) were more risk pants were asked to give a low and a high estimate that
seeking. Given the impact that individual risk attitudes they were 80% sure the answer lay between these numbers.
can have on corporate profits (Welsh and Begg, 2008), The first five of these questions were general knowledge
understanding how personality affects risk attitude is an with the following five being petroleum industry specific.
important observation, and part of this is the consideration Although half the questions were petroleum industry spe-
of how these traits affect biases related to risk attitude. cific, anyone should be able to answer these questions, as
For example, the framing effect (Tversky and Kahneman, greater uncertainty can be captured simply by increasing
1981) causes a significant proportion of people to reverse the width of one’s estimates range. These questions are
their choice between two identical options based solely used to test for overconfidence as follows: the expectation
on whether the choice is framed as a loss or a gain. Thus, is that if people are accurately setting 80% ranges, ~80%
we are interested to see whether people with particular of their ranges will contain the true value. If, however,
personality traits are more or less susceptible to this de- people’s ranges contain the true value less than 80% of
cision bias. the time then they are overconfident.
A question to measure the framing bias was adapted from
Hypotheses Tversky and Kahneman (1981). This question was framed
in both a positive and negative way, yielding a question
Given the above discussion, we are broadly interested pair. The question pair can then be used to determine if
in the question of whether personality traits affect sus- people change their risk preference or remain constantly
ceptibility to the particular biases under consideration risk averse or risk seeking. Each survey contained both
here. Given the details of that discussion, however, we members of the question pair to find whether individuals’
are in a position to make some more specific predictions responses to the questions changed when the question was
about how the MBTI types will relate to specific biases. described in terms of losses versus gains.

HYPOTHESIS 1 Procedure
MBTI N-type and F-type people will be more susceptible Both surveys were produced in both hard copy and on-
to all biases than S-types and T-types respectively. line via SurveyMonkey. Hard copies of the surveys were
distributed among petroleum engineering students at
HYPOTHESIS 2 the Australian School of Petroleum from first year up to
final year. After the hard copies had been distributed to
MBTI E-type people will be more overconfident and the students, a follow-up email was sent that reminded
more likely to be risk seeking than I-types. the students about the survey and also gave them the
opportunity to complete it online via SurveyMonkey for
METHOD those who had lost the hard copy. Between the hard copy
and online versions of the survey, about 50% of students
Participants completed the survey.
A number of petroleum exploration and production com-
Data was collected from 66 participants, including 38 panies were contacted via email through senior personnel
petroleum engineering students from the University of to explain the research project and included links to the
Adelaide and a further 28 participants, primarily (23 of electronic versions of the survey.
28) from the petroleum industry. Table 1 shows the break-
down of the participants by group, sex, age and industry RESULTS
experience.
Myers-Briggs
Materials
Before proceeding to analyses, it is important to know
A survey was designed consisting of two main sections: whether the four groups should be treated separately or
the first section being the MBTI; and the second containing together for statistical purposes—that is, whether they have
a series of questions to test for decision-making biases. the same characteristics or not. To assist in determining
The MBTI consisted of the 94 multiple-choice questions this, Table 2 shows the breakdown of each group along the
taken from Myers et al’s (1998) work. These questions were four MBTI type scales.
used to evaluate a person’s preferences on each of the four The engineer and other groups are both too small
dichotomies proposed by Myers and Briggs, and enabled to conduct meaningful statistics on, but to determine

SECOND PROOF—WELSH 10 FEB 11 APPEA Journal 2011—3


M. Welsh, A. Alhakim, F. Ball, J. Dunstan and S. Begg

Table 1. Group breakdown.

Group N Male Female Age (SD) Years experience (SD)

Student 38 27 10 21.6 (4.3) 0.6 (1.3)

Geoscientist 21 16 5 40.5 (9.7) 16.6 (9.3)

Engineer 2 2 0 23.5 (0.7) 1.5 (0.7)

Other 5 4 0 37.5 (18.0) 16.5 (18.6)

Note: two participants did not include their demographic details.

Table 2. MBTI types by group.

Group Total E/I S/N T/F J/P

Student 38 25/13 30/8 35/3 22/16

Geoscientist 21 12/9 12/9 19/2 12/9

Engineer 2 1/1 2/0 2/0 2/0

Other 5 2/3 2/3 4/1 4/1

Overall 66 40/26 46/20 60/6 40/26

whether the student and geoscientist groups should be do meaningful analyses of combination of types. For this
considered separately a series of chi-squared tests were reason, we restrict ourselves to comparison between the
conducted, comparing the proportion of each personal- paired types in each of the MBTI dichotomies.
ity types in these groups. These tests confirmed that the
proportions of the different types in the two groups did Overconfidence
not differ significantly, χ2(1) = 0.43, 3.13, 0.05 and 0.003,
for extraversion-introversion, sensing-intuition, thinking- A calibration score ranging from 0 to 10 was calculated
feeling and judging-perceiving, respectively, p >0.05 in all for each participant, from the 10 range estimation ques-
cases1. Given this, in all future analyses the entire data-set tions included in the survey. This score simply reflects
is considered together. the number of times that the ranges a participant stated
Looking at the overall pattern in Table 2, one sees cer- they were 80% confident would contain the true answer
tain types are more heavily represented in our sample to the question in, actually did. Figure 1 shows the mean
than others. Compared to the population estimates given calibration scores from each of the four pairs of traits
by Capretz (2003), our sample has slightly more extraverts and compares these to the expected calibration score of
than might be expected and also somewhat more judgers. 8—which would be the mean result if participants were
Sensors outnumber intuitors even more heavily, but this accurately setting 80% confidence intervals. Looking at
is in line with the overall population data. The greatest Figure 1, one sees in all but one case (the feelers), the
discrepancy is seen in the T/F dimension, where less than confidence intervals around the mean calibration scores
10% of our sample are feelers. In a random sample this do not include the expected value of 8. This indicates a
would be unexpected, as 50–60% of the total population significant difference between expected and observed
are Fs but previous research has shown a preponderance calibration. In all cases, the observed calibration is lower
of thinkers among engineers (McCaulley et al, 1983); the than the expected value of 8, indicating overconfidence.
fact that the same holds true for engineering students The fact that the confidence interval around the feelers’
and geoscientists is not too surprising. This does, however, mean calibration does include the expected value reflects
mean that statistical analyses carried out on the Fs will the very small sample size in this instance—only six indi-
be limited in their power and also restricts our ability to viduals—which acts to increase the width of confidence

4—APPEA Journal 2011 SECOND PROOF—WELSH 10 FEB 11


Do personality traits affect decision making ability: can MBTI type predict biases?

Figure 1. Mean calibration scores by MBTI type.

intervals as we are less sure of the reliability of our data. Framing


To determine how people’s calibration varied with the
personality types, a series of t-tests (Student, 1908) were Each participant saw two framing questions, forming a
undertaken, comparing the calibration scores achieved by matched pair.The first question of the pair was presented in
each type in the MBTI pairs. These indicated that, contrary a positive frame, where participants had to choose between
to the prediction of Hypothesis 2, extraverts were not more a certain $250 gain and a 25% chance of a $1,000 gain. The
overconfident than introverts. In fact, while not signifi- second also offered two alternatives to choose from with
cantly different, the extraverts’ calibration scores were expected values of $250, but framed the problem as a loss
slightly higher, indicating less overconfidence, t(62) = 0.42, by ‘giving’ the participant $1,000 to start the game and
p = 0.67 (two-tailed). then having them choose between a certain $750 loss and
The first prediction from Hypothesis 1, that feelers might a 75% chance of losing the entire $1,000. Participants were
be more susceptible to this bias than thinkers, was not sorted into one of four groups based on their pattern of
borne out by the data, with the calibration scores for the responses to these framing questions as shown in Table 3.
two groups being quite similar and, again, in the opposite Looking at Table 3, one sees that the classic framing
direction to our prediction—that is, feelers had higher effect is observed when a person chooses the risk-averse
calibration scores and were thus less overconfident, t(6) answer when a question is framed in a positive way (i.e.,
= 0.52, p = 0.62 (two-tailed). Similarly, the second Hypoth- in terms of gains) while preferring the risk-seeking option
esis 1 prediction, that sensors might be less susceptible to when the question is framed negatively (i.e., in terms of
overconfidence than intuitors, is exactly the opposite of losses). There are, however, three other possible response
what we see in the data where the intuitors have better patterns: a person can always be risk averse, always be
calibration and thus significantly lower overconfidence, risk seeking or do exactly the opposite of what the fram-
t(48) = 2.2, p = 0.03 (two-tailed). Finally, our hypotheses did ing effect literature predicts, being risk seeking for gains
not predict any difference between judgers and perceivers. and risk averse for losses. Table 4 shows the breakdown of
The data confirmed this, with the two types having simi- our participants into these four groups, divided according
lar calibration scores which the t-test indicated were not to their MBTI typology as well.
significantly different, t(54) = 0.79, p = 0.43 (two-tailed). Looking at Table 4, one sees that while there is a fram-
Overall, it seems the personality traits, as measured by ing effect it is not universal. Rather, it is the result of
the MBTI, do not predict susceptibility to the overconfi- about one-third of the participants acting as Tversky and
dence bias in the manner expected. The data do, however, Kahneman predict, with the majority of others having
support the idea that personality can affect susceptibility a firmer risk attitude—either risk-averse or risk-seeking
to bias, with the intuitors being significantly less overcon- regardless of the frame of the question. That is, there are
fident than the sensors on this task. Exactly why this might clear individual differences in the responses to framing.To
be, however, we defer until the discussion. determine whether these differences in framing behaviour

SECOND PROOF—WELSH 10 FEB 11 APPEA Journal 2011—5


M. Welsh, A. Alhakim, F. Ball, J. Dunstan and S. Begg

are related to participants’ personality traits, a series of than perceivers who are instead likely to be risk averse.
2x4 chi-squared tests were conducted, one for each of the Figure 2 shows the data from Table 4 converted into
personality dichotomies. That is, the proportion of people percentages to allow easy comparison between groups of
Es and Is using each of the four Framing behaviours were different sizes. Looking at this figure and recalling that
compared, and then the proportion of Ss and Ns, and so on. Hypothesis 1 suggested that N and F types would be more
These showed no significant differences between: ex- susceptible to biases, we generally see some evidence of
traverts and introverts, χ2(3) = 1.54, p>0.05; or sensors and this here, with 67% of (our very small number of) Fs suc-
intuitors, χ2(3) = 1.43, p>0.05. These tests did, however, cumbing to the framing effect, compared to only 30% of
indicate significant differences between the patterns of Ts. The difference between the percentage of Ss and Ns,
responding for judgers and perceivers, χ2(3) = 9.83, p<0.05; who display the framing effect (35.6% versus 28.6%, re-
and for thinkers and feelers, χ2(3) = 8.68, p<0.05—although spectively), however, is small and in the opposite direction
the very small numbers in the F cells call into question this to that predicted. There is also some support for the find-
result’s reliability. Looking at the data in Table 4, one can ings of Li and Liu, with a greater proportion of Ns than
see where these differences arise. Thinkers are far more Ss being risk seeking (28.2% versus 17.8%).
likely to be risk averse than feelers, who—if we believe Finally, Hypothesis 2 suggested that Es would be more
the current pattern will hold as a general trend—tend likely to be risk seeking than Is. While this is true, with
to be more often affected by framing. Similarly, judgers 22.5% of Es being risk seeking, almost as many ‘Is’ are
seem to be more likely to show the classic framing effect as well (19.2%). As such, there seems little support for
Hypothesis 2 in these data.
Table 3. Characterisation of participants by framing behav-
iour. DISCUSSION
MBTI E-type people will be more overconfident and
Negatively framed question more likely to be risk seeking than I-types.
Overall, the above results offer limited evidence to sup-
port the idea that personality traits—specifically those
measured by the MBTI—can be used to predict suscepti-
Risk averse Risk seeking bility to decision biases. A particular concern is while the
descriptions of the MBTI traits led us to posit particular
hypotheses regarding how those traits might affect bias
susceptibility, the majority of the significant effects we
observed were not those predicted. This casts doubt on
Positively framed question

the parallel drawn in the introduction between the MBTI


types and the Stanovich and West (2008) System 1/System
Risk averse Risk averse Framing
2 distinction. For example, while we found a number of
significant relationships between the personality traits
and the biases we studied, it does not seem to be the case
that intuitors and feelers are, as was originally suggested
in Hypothesis 1, more likely to use System 1 (intuitive
thinking) and thus be more prone to bias. In fact, on one
Risk seeking Reverse framing Risk seeking bias we found exactly the opposite of this, with intuitors
being significantly less overconfident than sensors—a re-
sult that needs some explanation.

Table 4. Breakdown of participants by MBTI and framing types.

E I S N T F J P

Framing 13 9 16 6 18 4 17 5

Risk averse 16 12 20 8 27 1 12 16

Risk seeking 9 5 8 6 14 0 10 4

Reverse framing 2 0 1 1 1 1 0 2

Total 40 26 45 21 60 6 39 27

6—APPEA Journal 2011 SECOND PROOF—WELSH 10 FEB 11


Do personality traits affect decision making ability: can MBTI type predict biases?

Figure 2. Percentage of each MBTI type showing specific framing responses.

A possible reason for this lies in the fact that accord- Caveats and future research
ing to the MBTI typology, intuitors prefer to make their
decisions intuitively—that is, without feeling a need to As always, there are a number of caveats to consider with
resolve uncertainties—whereas sensors prefer to use the respect to the above results. Some of these relate to the
information currently available to them and may as a result composition of our sample. As noted in the introduction, we
feel less comfortable making highly uncertain estimates. chose to look at personality traits rather than intelligence
If this is the case, then one might expect that the advan- due to the problem of restricted range caused by testing
tage that the intuitors have here is a result of the high people from an industry where a Bachelor’s degree is often
levels of uncertainty participants felt about the particular regarded as a minimum entry requirement. Comparing the
questions being asked and, as a result, if questions that distribution of personality types in our sample with those
the participants felt more knowledgeable about had been given for the overall population by Capretz (2003) however,
asked this difference might disappear. we see a similar problem may be occuring here. Feelers
The results on the framing task were equally confus- make up ~55% of the total population yet less than 10% of
ing in terms of our initial expectations. Our sample only our participants fell into this type. As a result, it becomes
partially replicated the results of Li and Liu (2008), with very difficult to know whether the distinctions between
more intuitors being risk seeking than sensors but we did thinkers and feelers that we have observed reflect genuine
not find the difference between extraverts and introverts population trends. It also makes it difficult to conduct
posited in Hypothesis 2 on either framing or overconfi- meaningful statistics comparing combinations of types
dence. Our strongest result from the framing task was with one another as the limited number of feelers would
the observation that judgers were far more likely than result in combinations of types that had no members at
perceivers to show the framing effect (i.e., changing their all. In fact, of the 16 possible types, only 13 were observed
risk attitude as a result of the question frame) but this was in our sample, with a further three being represented by
not a prediction we would have made based on the MBTI only one individual.
descriptions of these traits. Even considered in hindsight, Of course, given that our interest is in decision making
why these types differ in terms of framing makes little in the oil industry, we should expect some differences
sense, with the key distinction between the types being between the overall population distribution of types and
whether they like to have matters settled (judgers) or to the one that we observe. It may be that our sample is in
keep decisions open (perceivers). fact representative of oil and gas personnel. Data support-
We did, however, also have some evidence that feelers ing this is seen in Davey et al (1993), who found that 47%
were more likely than thinkers to show the framing effect. of mining personnel are ESTJ or ISTJ types—types that
This is in line with Hypothesis 1: that feelers might be account for 40% of our sample. If this is the case though,
more susceptible to the framing bias as a result of relying this limits the usefulness of the MBTI as a method for
on less rational decision-making methods. Given the very distinguishing between personnel and would imply that
small numbers of feelers in our sample, however, it seems future research might be better served by using an alter-
over-ambitious to hang too much on this finding. native personality test.

SECOND PROOF—WELSH 10 FEB 11 APPEA Journal 2011—7


M. Welsh, A. Alhakim, F. Ball, J. Dunstan and S. Begg

Such a move would probably have further advantages. contexts, it is important to keep in mind that the MBTI,
As noted in the introduction, our choice of the MBTI was according to its own creators, is designed to measure pref-
informed by the fact that it is well known and widely used erence, not aptitude. Thus, even accepting the underlying
in industry. It does, however, have a number of problems psychological theory on which the MBTI is based and that
that could be overcome by a change to another test. The a person’s MBTI type is accurate (both statements that
first is avoiding the false dichotomies represented by the a contingent of academic psychologists would debate),
MBTI types. As noted above, the MBTI on the basis of a this does not tell you whether they are good at making
series of questions designates people as either extraverts decisions in a particular manner—only that that is their
or introverts, assuming there is a sharp distinction between preferred manner.
these types. In fact, most people are somewhere in between Despite this, we still feel a more in-depth consideration
the two extremes and a significant number of people fall of the role of individual differences in decision making is
near the artificial dividing line and could, on any given day, necessary. As seen in our framing results, overly simplistic
be placed in either category. This undermines our ability characterisation such as ‘people are risk averse for gains’
to use the MBTI to distinguish between genuine extraverts miss a significant portion of what people are actually do-
and introverts as its E and I groups contain a large number ing. Given this, we feel it is important to continue trying
of people who are, in reality, neither. For this reason, it to determine what causes one person to be, for example,
may be better to move to a test like the NEO PI-R (Costa risk averse while another changes according to the way
and McCrae, 1992) that gives people a numerical score the question is framed.
on each of its dimensions rather than sorting them into
categories. Alternative tests also have more than the four ACKNOWLEDGEMENTS
dimensions considered by the MBTI—with the NEO PI-R,
for example, measuring five dimensions each comprised The authors wish to thank the following people for their
of a number of related facets. This finer-grained division assistance in contacting the industry personnel tested
herein: Peter Gilliland of EPM Training Services; Teof
of personality may enable differences to be isolated that
Rodrigues of Santos; Yacoub Al Dossary of Saudi Aramco;
the MBTI simply can not identify.
Ian Sylvester and Janny Spilsbur of Woodside; and Steve
Another potential concern might be that we have only
Mackie of Bass Oil.
tested a limited number of biases. While this is undoubt-
edly true, the biases examined here were chosen because
they seemed the most likely of those commonly consid- NOTES
ered to be affected by personality traits rather than being 1. Null hypothesis significance testing (NHST) is used
a result of intelligence-based memory or computational throughout. The p values given in the text represent
limitations.Thus, while the consideration of further biases the probability of finding as large a difference as
is a necessary next step, we do not believe that their inclu- was observed in the data by chance alone (that is,
sion here would in any way have altered our conclusions if there really is no difference between the groups
regarding the usefulness of the MBTI as a predictor of being compared except that due to sampling error).
the decision-making ability. By convention, a p value below 0.05 is regarded as in-
Finally, despite our concerns about the restricted range dicating a significant difference between the groups.
of intelligence we would encounter if testing oil and gas That is, if there is a less than 5% chance of seeing a
personnel, it may be worth considering a more detailed difference between groups as large as was observed
analysis—looking not at overall intelligence but rather at in the data by chance alone, we conclude the groups
the range of specific intelligences identified in the work of are significantly different from one another on the
Carroll (1993). As noted above in the discussion of alterna- variable being measured. A p value greater than 0.05,
tive personality tests, a finer-grained analysis may enable conversely, results in our conclusion that the groups
distinctions to be drawn between the types of intelligence do not differ from one another. The 95% confidence
that help render one person resistant to biases. intervals displayed on Figure 1 show this same statisti-
cal effect pictorially.
CONCLUSIONS
REFERENCES
In conclusion, while we found a small number of rela-
tionships between MBTI types and performance on the BLOCK, R.A., AND HARPER, D.R., 1991—overconfidence
overconfidence and framing tasks used herein, these results in estimation: testing the anchoring-and-adjustment hy-
do not paint a consistent picture about the type of person pothesis. Organizational Behavior and Human Decision
one should prefer to make decisions, with less susceptibility Processes, 49 (2), 188–207.
to bias.This is of particular importance as the relationships
we found were not those that we expected to find based CAPRETZ, L.F., 2003—Personality types in software
on our reading of the MBTI materials. That is, while the engineering. International Journal of Human-Computer
descriptions of specific MBTI types may seem to indicate Studies, 58 (2), 207–14.
a person will have strengths in particular decision-making

8—APPEA Journal 2011 SECOND PROOF—WELSH 10 FEB 11


Do personality traits affect decision making ability: can MBTI type predict biases?

CARROLL, J.B., 1993—Human cognitive abilities: a NORTHCRAFT, G.B. AND NEALE, M.A., 1987—Experts,
survey of factor-analytic studies. Cambridge: Cambridge amateurs and real estate: an anchoring-and-adjustment
University Press. perspective on property pricing decisions. Organizational
Behavior and Human Decision Processes, 39 (1), 84–97.
COSTA, P.T.J. AND MCCRAE, R.R., 1992—NEO PI-R
Professional Manual. Odessa: Psychological Assessment RUSSO, E.J. AND SCHOEMAKER, P.J.H., 1992—Managing
Resources. Overconfidence. Sloan Management Review, 33 (2), 7–17.

DAVEY, J.A., SCHELL, B.H. AND MORRISON, K., 1993— STANOVICH, K.E. AND WEST, R.F., 1998—Individual
The Myers-Briggs Personality Indicator and its usefulness differences in rational thought. Journal of Experimental
for problem solving by mining industry personnel. Group Psychology: General, 127 (2), 161–88.
and Organization Management, 18 (1), 50–65.
STANOVICH, K.E. AND WEST, R.F., 2008—On the relative
H AW K I N S , J. T. , C O O P E R S M I T H , E . M . A N D independence of thinking biases and cognitive ability.
CUNNINGHAM, P.C., 2002—Improving stochastic evalu- Journal of Personality and Social Psychology, 94 (4), 672–95.
ations using objective data analysis and expert interview-
ing techniques. SPE Annual Technical Conference and STUDENT, 1908—The probable error of a mean. Biometri-
Exhibition, San Antonio, Texas, 29 September–2 October, ka, 6 (1), 1–25.
SPE 77421.
TVERSKY, A. AND KAHNEMAN, D., 1974—Judgment
JUNG, C.G., 1971—Psychological Types (Collected Works under uncertainty: Heuristics and biases. Science, 185,
of C.G. Jung, Vol. 6. Princeton, New Jersey: Princeton 1124–31.
University Press.
TVERSKY, A. AND KAHNEMAN, D., 1981—The framing of
JUSLIN, P., WENNERHOLM, P. AND OLSSON, H., 1999— decisions and the psychology of choice. Science, 211, 453–8.
Format dependence in subjective probability calibration.
Journal of Experimental Psychology: Learning, Memory WELSH, M.B. AND BEGG, S.H., 2008—Modeling the Eco-
and Cognition, 25 (4), 1038–52. nomic Impact of Individual and Corporate Risk Attitude.
SPE Annual Technical Conference and Exhibition, Denver,
LI, S. AND LIU, C.J., 2008—Individual differences in a Colorado, 21–24 September, SPE 116610.
switch from risk-averse preferences for gains to risk-seeking
preferences for losses: can personality variables predict the WELSH, M.B., BEGG, S.H. AND BRATVOLD, R.B., 2007—
risk preference? Journal of Risk Research, 11 (5), 673–86. Modeling the economic impact of cognitive biases on oil
and gas decisions. SPE Annual Technical Conference
LICHTENSTEIN, S., FISCHHOFF, B. AND PHILLIPS, and Exhibition, Anaheim, California, 11–14 November,
L.D., 1982—Calibration of probabilities: the state of the SPE 110765.
art to 1980. In: Kahneman, D., Slovic, P. and Tversky, A.
(eds.)Judgment under Uncertainty: Heuristics and biases. WELSH, M.B., BEGG, S.H., BRATVOLD, R.B. AND LEE,
Cambridge: Cambridge University Press. M.D., 2004—Problems with the Elicitation of Uncertainty.
SPE Annual Technical Conference and Exhibition, Hous-
MCCAULLEY, M.H., GODLESKI, E.S., YOKOMOTO, C.F., ton, Texas, 26–29 September, SPE 90338.
HARRISBURGER, L. AND SLOAN, E.D., 1983—Applica-
tions of psychological type in engineering education. WELSH, M.B., BRATVOLD, R.B. AND BEGG, S.H., 2005—
Engineering Education, 73 (5), 394–400. Cognitive Biases in the Petroleum Industry: Impact and
Remediation. SPE Annual Technical Conference and Ex-
MCCRAE, R.R. AND COSTA, P.T., 1989—Reinterpreting hibition, Dallas, Texas, 9–12 October, SPE 96423.
the Myers-Briggs Type Indicator from the perspective of
the five-factor model of personality. Journal of Personality WINMAN, A., HANSSON, P. AND JUSLIN, P., 2004—Subjec-
and Social Psychology, 57 (1), 17–40. tive pro bability intervals: how to reduce overconfidence by
interval evaluation. Journal of Experimental Psychology:
MORGAN, M. G.,AND HENRION, M., 1990—Uncertainty: A Learning, Memory and Cognition, 30 (6), 1167–75.
Guide to Dealing with Uncertainty in Quantitative Risk and
Policy Analysis. Cambridge: Cambridge University Press.

MYERS, I.B., MCCAULLEY, M.H., QUENK, N.L. AND


HAMMER, A.L., 1998—MBTI Manual (A guide to the
development and use of the Myers Briggs Type Indicator).
Palo Alto: Consulting Psychologists Press.

SECOND PROOF—WELSH 10 FEB 11 APPEA Journal 2011—9


M. Welsh, A. Alhakim, F. Ball, J. Dunstan and S. Begg

THE AUTHORS

Matthew B. Welsh has a BA (Hons) in Joseph Dunstan completed his BPetEng


philosophy and a BSc (Hons) and PhD and BCivilEng at the University of
in psychology, all from the University of Adelaide in 2010. He has worked on
Adelaide. He has been employed as a civil infrastructure projects and is now
research fellow in the Improved Business employed as a subsea engineer with
Performance Group at the Australian Woodside. Member: SPE, Engineers
School of Petroleum for the past seven Australia.
years, conducting research focusing on the
psychological aspects of decision making
and the impact these can have on the accuracy of forecasts Stephen (Steve) H. Begg has a BSc and
and thus economic outcomes. Member: SPE, Cognitive Science PhD in geophysics from the University of
Society, Association for Psychological Science. Reading, England. He is now is a Professor
of Petroleum Engineering and Management
Abdulaziz Alhakim completed his at the University of Adelaide, Australia,
Bachelor of Petroleum Engineering (Hons) focusing on decision making under un-
at the Australian School of Petroleum in certainty, asset and portfolio economic
2010. He was sponsored by, and is now evaluations, and psychological factors that
employed by, Saudi Aramco as a drilling impact these. Formerly, he was the Direc-
engineer. Member: SPE. tor of Strategic Planning and Decision Science with Landmark
Graphics. Prior to that he held a variety of senior operational
engineering and geo-science roles for BP Exploration, and was
a reservoir characterisation researcher and manager for BP
Finlay Ball completed a combined degree Research. Steve has been an SPE distinguished lecturer and has
in BEng (Petroleum) and BEng (Civil and chaired several SPE Forums and advanced technology workshops
Structural) in 2010. He has been employed related to his areas of expertise. Member: SPE, AAPG.
by ShellAustralia since March 2011 as a civil
engineer. Member: SPE, PESA, Engineers
Australia.

10—APPEA Journal 2011 SECOND PROOF—WELSH 10 FEB 11


View publication stats

You might also like