Professional Documents
Culture Documents
DAVID LAMOND
Macquarie University
email: david.lamond@mq.edu.au
The Myers-Briggs Type Indicator:
Evidence Of Its Validity, Reliability And Normative Characteristics For Managers In An
Australian Context
ABSTRACT
The Myers-Briggs Type Indicator (MBTI) is widely used as a selection and training and
development tool in Australian organisations. While a great deal of accessible information for
the American samples on which it was developed is available, to date, there has been little in
the way of published data to guide its use in an Australian context. The purpose of this paper
is to review the available literature concerning the reliability and validity of the MBTI and to
present the results of the administration of the instrument to a group of 523 Australian
managers, as a basis for establishing an appropriate set of relevant norms. The results indicate
that the MBTI is a valid and reliable instrument in an Australian organisational context. At
the same time, it is argued that, while the MBTI is a valid and reliable instrument for this
group, its use should continue to be carefully circumscribed, especially in the absence of
Psychologists and academics have advocated the use of personality measures as part of the
decision-making battery for managerial selection for many decades (Dissanayake 1993:9).
However, psychological tests (cognitive and personality tests) were, until recent times, used
only irregularly (Thomas 1993:96). In Britain during the 1980s, for example, personality tests
were only used by a third of companies, and in those companies which did use them, 70% did
so in less than half their selection processes (Robertson and Makin, 1986:47). Similar
patterns appeared in Australia (Patrickson and Haydon, 1988; Vaughn and Maclean, 1990)
and in the South-East Asian republic of Singapore (Chew and Yeo, 1991).
Coincidentally, this sporadic use of personality tests reflects the doubts of Guion’s and
Gottier’s (1965:160) conclusion that “it is difficult in the face of this summary to advocate,
with a clear conscience, the use of personality measures in most situations as a basis for
making employment decisions”. Criticism about the use of personality measures as part of
the personnel selection process has continued since that time (see, for example, Fletcher,
Blinkhorn, and Johnson, 1991; Guion, 1991; Spillane, 1994), and writers continue to urge
caution in the use of personality measures for personnel selection (for example, Macy, 1994).
Indeed, the search for personality correlates and determinants of occupational behaviour has
been described as akin to the search for the Holy Grail - it has been long standing, replete with
myths and legends and, to date, largely unsuccessful (Furnham and Stringfield, 1993:827-
828). Nonetheless, researchers and practitioners continue to try to identify these relationships
and, in more recent times, this pessimism has been tempered considerably (see, for example,
In regard to explaining and predicting managerial behaviour, the two main sets of personality
measures in use are those that measure the “Big Five” factors of personality (cf Barrick and
1
Mount, 1991) and the Myers-Briggs Type Indicator (MBTI) (Myers and McCaulley, 1985).
These two approaches represent fundamentally different philosophies about personality, the
former firmly rooted in a trait-based pedigree (Digman, 1990; Costa and McCrae, 1993) and
the latter a type-based product of analytical psychology, informed by the work of Carl Jung
(1971). The trait or dimensional approach relies on correlational techniques to uncover linear
relationships between behaviours across a population, and so reveal the sources of behaviour.
While the MBTI is widely used in Australia, there is little in the way of validity, reliability
and normative data to inform its use with managers in an Australian context. The purpose of
this paper is to present the results of a study that included the administration of the MBTI to a
sample of more than 500 Australian managers. The results provide evidence of the validity
and reliability of the MBTI for this group, but also point to differences in the distribution of
type amongst Australian managers when compared to managers in the US and UK. The paper
begins by describing the MBTI and the current literature on its reliability and validity. The
methodology for the current study is then explained, with emphasis given to presenting details
about the sample so that the reader is able to draw independent conclusions about the veracity
of the resultant normative information. The implications of the results for the use of the
which is used to measure Jung’s psychological types (Myers, 1962; Myers and McCaulley,
1985; Myers, McCaulley, Quenk and Hammer, 1998). The MBTI was developed initially in
2
the late 1940s, as a way of operationalizing that part of Jung's overall theory which is
specifically concerned with the systematic differences in the way individuals consciously
prefer to use their perception and judgement, and the corresponding differences in their
interests, reactions, values motivations, and skills (Myers, et al, 1998:3). There are 16
possible types, derived from measures of four dichotomies: Extraversion (E) - Introversion
(I); Sensing (S) - Intuition (N); Thinking (T) - Feeling (F); and Judgement (J) - Perception (P).
adjectives. Each is designed to elicit a preference for one of the functions, attitudes or
orientations, with a score of 0, 1 or 2 given to the answer to indicate the weight of preference.
These scores are tallied to give a weighting for each of the eight preferences. These, in turn,
are converted to a preference score along each of the four dimensions, reflecting the
preference for one pole over the other (cf Myers and McCaulley, 1985:9). A brief account of
the eight preferences is presented in Table 1. Taken together, the four preferences indicate a
person’s MBTI type. For example, a person with a preference for extraversion (E) over
introversion (I), for intuition (N) over sensing (S), for thinking (T) over feeling (F), and for
The MBTI has become the most widely used personality questionnaire in America for non-
psychiatric populations (Murray, 1990) and more than 4,000 research studies, journal articles
and dissertations have been written about the MBTI since 1962, producing a great deal of
analytical and normative data (Myers, et al, 1998:10). Proponents argue that the data
generated about individuals and groups from the MBTI (McCaulley, 1990a) are useful in
vocational and other forms of counselling (eg, McCaulley, 1990b); in learning and teaching
3
(eg, Cooper and Miller, 1991); and in refining the understanding of leadership styles (eg,
McCaulley, 1990c).
Distribution of the various preferences in the Australian population has been difficult to
establish, since no large random sample of the population has ever been undertaken. Myers
and McCaulley (1985:40) presented the type profile of a group of Australian MBA students,
while, more recently, Myers, et al (1998:379) reported the results of a survey of Australian
high school students. However, neither constituted a random, representative sample of the
Australian population. Power (1989) has reported the results of approximately 2,400 MBTI
workshop participants, but the data are so biased as to be not particularly useful, being records
collected mostly from individuals who had an affiliation with the Roman Catholic Church,
and who lived in Brisbane or in provincial areas of Victoria. None lived in Sydney.
Myers, et al, (1998:379) present type distribution for national normative samples of adults in
the US (n = 3,009) and the UK (n = 1,634). An adapted version of this table is presented in
Table 2 to show the distribution of type in two countries which are culturally similar to
Australia (cf Hofstede, 1980; 1991). Chance distribution of 16 types in the 4 x 4 matrix in
Table 2 would produce 6.25% per cell, and so casual inspection suggests a non-random
distribution of types. At the same time, Table 2 also shows that, even where countries share
similar cultures, there is some variation in the distribution of type. Myers, et al (1998:384)
caution about use of the inventory on people culturally different from those with whom it was
developed.
Despite the MBTI being administered to more than three million people each year (Spoto,
1995), it continues to be the subject of extensive criticism in the literature (cf Bayne, 1995;
4
Boyle, 1995; Carlson, 1985, 1989; Healy, 1989; Lorr, 1991; and Stricker and Ross, 1963,
1964). At the same time, there has been a plethora of research to examine the validity and
reliability of the MBTI, and generally the data support the view that the MBTI scales are valid
(see, for example, Murray, 1990; Thompson and Borrello, 1986; Tzeng, Ware, and Bharadwaj,
1991; Wiggins, 1989) and reliable (see, for example, Howes and Carskadon, 1979;
McCaulley, 1990a; Myers and McCaulley, 1985; Pearman, 1993). However, a problem
MBTI data have been gathered. The result is that it is difficult to summarise conveniently the
literature in order to make a case for its validity and reliability. The following sections will
Myers and McCaulley (1985) point to an abundance of data which demonstrate that
preferences about treatment modalities and learning choices consistent with type expectations,
as indicative of the construct validity of the MBTI. Support for the construct validity of the
MBTI is mixed, but the majority of the evidence is consistent with the view that the MBTI
does measure important dimensions of personality that approximate those in Jung’s typology.
Myers and McCaulley (1985:176-208) report a series of correlations between the MBTI
scales and personality characteristics measured by other questionnaires and selected on the
basis that, according to MBTI theory, the questionnaires would be measuring either the same
construct as one of the scales or one that should be associated with a scale. The correlations
range from –0.77 to 0.75, varying according to the scale and characteristics, but they are all
statistically significant and in the expected direction. Other reports show similar and even
5
higher correlations (see Brown and DeCoster, 1991; Murray, 1990; Myers, 1962; Steele and
In regard to item factor structure, there have been conflicting reports. Tzeng, Outcault, Boyer,
Ware, and Landis, (1984) report a four factor structure matching the MBTI scales, but Sipps,
Alexander and Friedt (1985) report a six factor structure, with four equating to the MBTI
scales. Using a second-order factor analytic technique, Thompson and Borrello’s (1986)
analysis of data from 359 college students arrived at a four-factor solution, with only three
test items loading on incorrect factors, and so lending support for the construct validity of the
four MBTI scales. In a later study by Sipps and DiCaudo (1988), analysis of the responses of
185 psychology students to the EI and JP subscales supported the EI scale as a measure of
Linked to the variable clusters, is the extent to which each of the scales is relatively
independent of each other. This has generally been found to be the case except for the SN and
JP scales. Myers (1962) reported a correlational analysis showing EI, SN, and TF scales to be
independent, while there was a correlation between the SN and JP scales. McCaulley (1977)
also reported a significant correlation between the SN and JP scales. Myers and McCaulley
a fact about the types themselves. Sensing types typically prefer to rely on past experience,
and dislike unexpected events that require rapid assessment of new possibilities … Intuitive
types, on the other hand, are attracted to future possibilities and new construction of events
Looking at the distribution of types in the US and UK populations given in Table 2, this
seems to be the case, with SJ (47%) and SP (26%), while NJ (8%) and NP (18%).
Croom, Wallace, and Schuerger, (1989) have been able to develop regression equations to
predict MBTI scores from Cattell’s 16PF Questionnaire scores, based on intercorrelations
6
given in the handbooks for the two instruments. These were cross-validated using data from
two samples of varying populations (students and managers). In terms of ability to predict
both single preference scores and the 4-letter type, accuracy was found to approximate the
short-term test-retest reliability of the MBTI. These findings provide further evidence of the
McCrae and Costa (1989) compared the scores of 468 respondents on the MBTI and the
NEO-PI and found positive correlations between four of the five NEO-PI scales and the four
MBTI scales, viz: Extroversion and Extraversion, Openness and Intuition, Agreeableness and
Feeling, and Conscientiousness and Judging, in the range 0.44 to 0.74. There was, of course,
no correlation between the NEO-PI Neuroticism scale and the MBTI scales. McCrae and
Costa (1989) formed the view, that there was no support for the view that the MBTI measures
truly dichotomous preferences or qualitatively distinct types; instead, they argued that the
instrument measures four relatively independent dimensions. They believe their data suggest
that Jung's theory is either incorrectly or inadequately operationalized by the MBTI and
cannot provide a sound basis for interpreting it. They argue that the 5-factor model provides
an alternative basis for interpreting MBTI findings within a broader, more commonly shared
conceptual framework. This view ignores the fact that the findings reported by McCrae and
Costa (1989) are consistent with expectations of both trait and type theorists. One would
expect, for example, an Extraverted, Intuitive, Feeling, Judging (ENFJ) type to score more
highly on the extraversion scale (by definition if nothing else), and on the Openness to
McCrae and Costa (1989) found. This appears to offer evidence of convergent validity.
Carlyn (1977) concludes, and the studies examined above appear to support the view that,
“the individual scales of the Myers-Briggs Type Indicator measure important dimensions of
7
personality which seem to be quite similar to those postulated by Jung. The indicator appears
The reliability of the MBTI as a measure of Jung’s typology has been established in a number
of the conventional ways. In regard to the internal consistency of the MBTI, for example,
conclude that they are “consistent with those of other personality instruments, many of which
have longer scales than the MBTI”. The internal consistency has also been measured by way
of coefficient alphas (Cronbach, 1951), where scores for the continuous scales have been
In an early examination of the reliability of the MBTI, Stricker and Ross (1964) found the EI,
SN, and JP scales to have good internal consistency reliability (α = .75 - .85) and retest
stability (α = .70). The TF scale, however, showed low internal consistency (α = .44) and
stability (α = .48). Myers (1980) maintains that this lower retest reliability on the TF scale is
a function of the lower maturity of the sample group that was used to test the measure. This
view is consistent with type development theory, and found support from McCaulley (1977)
who showed in a summary of internal consistency reliability studies of the TF scale that the α
ranged from 0.44 in 8th grade to 0.8+ in high school seniors and college students.
It was crucial to establish test-retest reliability, as a measure of the stability of the MBTI, for
one principal reason. Both Jung (1921/1971) and the developers of the MBTI (Myers and
McCaulley, 1985; Myers et al, 1998) saw personality type as invariant. If this is the case,
then test-retest reliabilities should be high. Again, Myers and McCaulley (1985:170-174)
report studies where the between-test period has ranged from four to five weeks to five years.
Reliability is consistently good, with correlations between continuous scores mostly in the
8
range 0.77 to 0.89. After a five-week period, agreement for preferences ranged from 77% to
92%, with approximately half the sample remaining the same on all four scales, a result
significantly better than chance (Myers and McCaulley, 1992:173). Even when the mood of
the respondents was deliberately elevated or depressed prior to the second administration,
results were consistent with those already mentioned (Howes and Carskadon, 1979). As
McCaulley (1990a:187) says, in about three-quarters of the cases, the retest will show the
same three or four letters, with changes occurring when the original preference score was low.
Most reviewers of the MBTI see these as positive results, being consistent with the test-retest
reliabilities of trait tests, such as the NEO-PI (Costa and McCrae, 1985, 1992). For example,
Pearman (1993:67) observes that the MBTI reliability studies show that it is more consistent
than either the California Psychological Inventory or the Adjective Check List, both of which
(1993:472) argues this is a poor result, since the MBTI is not a trait theory and “implies a
dichotomy where the classes are mutually exclusive … The reliability data suggest that an
accurate and durable assessment of type cannot be made for those subjects whose scores are
close to the zero point and have a high probability of crossing that boundary”. This actually
fits well with Jung’s (1921/1971:549) theory, since he did not believe that everyone could be
There must be two fundamentally different general attitudes which would divide human
beings into two groups – provided the whole of humanity consisted of highly differentiated
individuals. Since this is obviously not the case, one can only say that this difference of
attitude becomes plainly observable only when we are confronted with a comparatively
well-differentiated personality.
Jung’s caveat, of course, is that the identification of the two groups assumes that “the whole
516) argues, there is a third group, the least differentiated and most numerous, where it is hard
9
to say whether they are introverted or extraverted, and only careful observation will ensure a
correct classification. In the same way, the developers of the MBTI do not claim the MBTI
will always provide an accurate classification of type. Rather, “no questions, however
accurate, can explain all human complexity. The MBTI results are a first step toward
Notwithstanding Carlson’s (1989:485) concern about the haphazard patterns of the reports
about the validity and reliability of the MBTI, one of the many reviews of research on the
MBTI, has concluded that the MBTI’s indices of reliability and validity have been judged
acceptable, and the constructs underlying the MBTI have been supported by correlations with
other measures (Murray, 1990). Indeed, “[e]ven the most serious critics of the MBTI
recognize the validity of the instrument and acknowledge that it can be an effective
appropriate for use in team building, career development, organization development and
leadership development (Consulting Psychologists Press, 1994:7) and its use for these
purposes has not gone unnoticed. Coe (1992) remarks on the increasing use of MBTI for
dysfunction and for management selection and development. The broad claim of the sponsors
of the MBTI in management selection is fairly simple - “matching the personality with the
This “personality-centred approach to management” (Isachsen & Berens, 1989) argues that
management style - “the way in which an individual determines what needs to be done and
then create (sic) the necessary conditions to achieve the desired results” (Isachsen & Berens
10
1989:118) - is a function of personality and so differences in personality mean differences in
the ways in which people manage. In other words, if you know a person’s personality type,
you will also know that person’s management style. In order to increase the chances of
selecting the manager with the “right” style, one need only determine the person’s personality
In an organizational context, the MBTI is used inter alia, as a basis for explaining individual
differences in occupational choice (eg, Myers, 1980), ‘leadership’ style (eg, Barr and Barr,
1989; Keirsey and Bates, 1984), ‘management’ style (eg Benfari, 1991, 1995; Isachsen and
Berens, 1989) and managerial style (Margerison and Lewis, 1981). Again, it is claimed that
managers’ styles (ways of planning, organising, leading, coordinating and evaluating) will
vary according to their personality type. In general terms however, there is a paucity of type
research into managerial behaviour, matched by a general lack of rigour in the research that
In a wide ranging review of the research on type and managerial behaviour, Gardner and
Martinko (1996) summarise what research has been undertaken to date in terms of four
design. In light of the research on MBTI type in general, the research results described by
Gardner and Martinko (1996:59-77) are utterly unsurprising. First, they show that TJs are
over-represented in the manager group (cf Myers and McCaulley, 1985:40; Myers, et al,
1998:299; see also Table 3 below). Second they show that the characterisation of manager Ss,
Ns, Ts, Fs, Es, Is, Js and Ps is identical to that of the general population. Importantly, the
results highlight the fact that managers tend to report themselves as behaving according to
type in the work context. For example, Judging managers behave in a planned, conscientious
11
and methodical manner, while Perceivers are more spontaneous, flexible and creative
Two studies not reviewed by Gardner and Martinko (1996), one English and one from Hong
Kong, are worthy of discussion here and they are considered in turn. Taken together, these
latter studies highlight the importance of ensuring that any analysis is cognisant of country
and cultural differences amongst the respondents, as these are likely to influence the direction
An early attempt to examine the relationship between personality type and managerial style is
that of Margerison and Lewis (1981:3) who proffered what they claimed to be “Jung’s theory
organizations.” They presented a type distribution for 849 UK managers attending business
school short courses, which is compared with the UK national distribution of type in Table 3.
Examination of Table 3 clearly shows that not only are certain types of people attracted to
managerial roles, but also that certain types are selected more often.
Margerison and Lewis (1981:17) note that 52.8% of their managers are STs and so, they say,
these managers “will be concerned first and foremost with practical and logical problems.
They will also prefer problems that are concrete and specific rather than ambiguous and
abstract and, hence, their impatience and distrust of issues that to them seem nebulous and not
based on tangible factors”. Margerison and Lewis (1981:17) also note that those concerned
with people issues – SFs (15%) and NFs (10%) – are outnumbered 3:1 by those most
concerned with technical problems – STs (53%) and NTs (22%) – and suggest that this
perhaps “explains the need for ‘interpersonal’ skill courses for managers”.
12
More recently, Furnham and Stringfield (1993) compared the MBTI scores of a group of
Chinese (n = 222) and European (n = 148) managers working for a successful international
airline based in South East Asia, with behavioural ratings of their managerial practices (eg,
decision-making, planning and innovation). Furnham and Stringfield (1993) do not report
their findings by way of the traditional ‘type table’ (see, for example, Tables 3.2 and 3.3) so
one is not able to judge whether the type distributions for these two groups are similar to other
manager groups. Using continuous scale scores, Furnham and Stringfield (1993:837) report
significant differences between the Chinese and European groups: compared to the Chinese
sample, the Europeans were significantly more extraverted, more intuitive, more feeling, and
more perceiving. The average profile for the Chinese sample was ISTJ, while for the
In light of the reliability and validity findings outlined above, together with the evidence
suggesting sector, culture and country differences in type distribution, the aim of the current
study was to see whether the MBTI is a valid and reliable measure when used with managers
in Australia. A further aim was to report the MBTI type distribution as normative data, and so
METHODOLOGY
The results reported here are part of a larger research study which involved asking members
of the alumni of a prominent NSW business school, resident in Australia, to complete a postal
managerial behaviour (Author, 2000). The nature of the main research questions required a
large sample size for meaningful comparison and contrast of individual perceptions, as a basis
for reaching generalisable conclusions. The alumni group, at over 2,700 individuals, provided
a useful sample group, but it simply would not have been possible to administer the survey on
13
an individual, face-to-face basis or bring the individuals together in a series of locations for
group administration. Notwithstanding problems associated with low response rates, the mail
The survey group (managers with management qualifications) was, in turn, chosen for two
reasons. First, as Fayol (1949) points out, while managerial functions are not limited to those
who are called ‘manager’, the more a person is embedded in the management cadre, the more
s/he will engage in managerial rather than other technical organizational functions. Since the
focus of the wider research was on managerial behaviour, it was appropriate to sample from a
group of people who are more likely to have wide experience and engagement in the range of
consideration.
The second reason for choosing this group for the respondent sample relates to a theoretical
qualification may have on both preferred and enacted managerial style. The literature on
management education (cf Bilimoria, 1998; French and Grey, 1996; Porter and McKibbin,
1988) has not directly considered the impact of the experience on individuals’ management
style (at least as it is defined in this research). At the same time, exhortations to ‘improve’
management education (cf. Anon, 1991; Bilimoria, 1998; Commonwealth of Australia, 1995;
Das, 1994; French and Grey, 1996) certainly suggest that different management development
this is the case, then having a sample in which all its members possess a management
qualification should eliminate the variability associated with the possession (or not) of a
management qualification.
14
It is not enough though, to use the simple marker of a postgraduate management qualification.
A search of the Internet sites of the more than 40 business schools in Australia teaching at
postgraduate level reveals a diversity of philosophy, course contents and teaching and learning
resources. The business school whose alumni group was involved in the research is no
education, with its emphasis on the social and organizational context of management and its
use of syndicate groups as a key learning process (More, 1999). If the management learning
one business school therefore reduces the variability that would result from differences in the
learning experience.
The full research instrument was a paper-and-pencil survey questionnaire, which comprised
the Myers-Briggs Type Indicator, a Managerial Style Measure developed by the author for the
wider research program, Rotter’s (1966) locus of control measure, Quinn’s (1988) Competing
individuals’ biodata. The presentation order of Myers-Briggs Type Indicator and the
Managerial Style Measure was reversed for half of the questionnaires distributed, in order to
compensate for any order of presentation effects, but the forms were otherwise identical.
The specific version of the MBTI was ‘Form G’, a 94-item forced-choice questionnaire that
can be hand scored by the respondent or the administrator. The 94 items comprise 49 sets of
behavioural statements (all of which have two alternatives, except for three statements that
have 3 alternatives each) and 45 pairs of self-descriptive adjectives. Form G was selected
15
therefore appropriate for inclusion as part of a written survey. One might note that much of
the previously cited validity data originates from this form, as part of a written survey.
Respondents were told that the MBTI was the personality measure employed. This was done
because, as graduates of the particular business school and practising managers, they had all
been exposed to the MBTI in a variety of contexts. They were asked to read the directions
carefully then complete all questions. They were reminded there were no "right" or "wrong"
answers to the questions; rather, their answers would help to show how they like to look at
things and how they like to go about deciding things. The respondents were asked to read
each question carefully and circle the letter that corresponded to their answer, not to not think
too long about any question, and, where they felt more than one answer described them, to
A total of 2733 questionnaire packets were mailed to those alumni resident in Australia at the
time of the study, according to the School’s maintained alumni database. The packet
and signed covering letter seeking the respondents’ cooperation, an ethics consent form, and a
reply paid addressed envelope in which to return the completed instrument and consent form.
Recognising the difficulties associated with mail-out surveys in terms of their response rates,
efforts were taken to maximise the response rate for this survey. For example, each letter was
individually addressed in School envelopes, and respondents’ status as School alumni was
referred to in seeking their cooperation. The use of reply paid, pre-addressed envelopes for
return of the survey was to maximise the chances that the survey was returned, and returned to
the correct address. As an added incentive, respondents were advised that they would receive
an Executive Summary of the key findings, along with an individual profile of their survey
responses.
16
A total of 523 responses were received (241 Form A responses and 282 Form B responses).
This represented a response rate of 19.5%. Once the surveys were returned and catalogued,
they were forwarded to a data entry house for ‘double keyed’ data entry according to an SPSS
statistical software protocol developed by the author. When the data were keyed, a
subsequent check and data cleaning process were carried out as an extra precaution to ensure
the integrity of the data base. First, a random check of the completed surveys against the
entered data was completed. This showed an extremely high level of accuracy. A frequency
check of the data, using the SPSS software package identified a small number of mis-keyed
data. Those variations between the keyed data and the expected data were checked with the
original questionnaires to determine the correct value and the data file was adjusted
accordingly.
The following series of tables indicate the nature of the sample that provided responses. Table
4 summarises the demographic data for the respondent group, revealing nothing surprising.
For example, that three-quarters of the respondents are male reflects the enrolment patterns of
the School over its three decades of existence and, one might note, the past and current trends
for employment of women in management positions in Australia. The modal age group of the
respondents is 40-49 years. Again, this reflects the School’s practice of selecting experienced
Table 4 also shows that most respondents were born in Australia. Of the 15% of respondents
born in a non-English speaking (NES) country, the largest sub-group is Asian (40% of NES
group). Finally, the table shows that most (60%) respondents have an MBA/DBA as their
highest management qualification, and that most occupy a middle (32%) or senior
17
had small numbers of respondents and so were combined. Even then, the respondents in this
combined category represent only a small percentage (6%) of the total respondent group.
Taking all these characteristics into account and comparing them with the School alumni
group as a whole, it is reasonable to state that the sample here is representative of the wider
School alumni population. In order to establish whether there were systematic variations in
the sample demographic and organizational variables, a correlation analysis was carried out.
Table 5 shows that there are several small but significant correlations between sex and age
group (more males in the older age groups), management qualifications (males have higher
management qualifications), and sector (females are more likely to work in the public sector).
None of these findings is surprising or unusual. The older respondents (and therefore earlier
graduates) are more likely to be male. With the development of the Postgraduate Diplomas
and the Masters of Management have come more female graduates (who may not have the
undifferentiated, qualification). The correlation between age and level in the organisation
(older people hold higher positions) is expected. There are no significant relationships
Based on the results of the correlation analysis, the nature of the identified relationships was
further explored, by way of χ2 analysis, the significant results of which are summarised here
in Table 6. Individually, these results are only of passing interest. Taken together, however,
they do highlight significant, systematic differences in the respondent group, related to sex
18
and age. As such, they provide insights that may afford important detail for interpreting
subsequent findings.
RESULTS
It was noted earlier that two forms of the research instrument were used – Form A presented
the Myers-Briggs Type Indicator first and Form B presented the Managerial Style Measure
first. This was done to compensate for any order of presentation effects. The final data set
contained 241 responses from Form A and 282 responses from Form B. To test whether the
ordering of the questions had affected the respondents’ answers to the MBTI, the reliability of
the MBTI was determined separately for the two forms. Before doing so, some re-coding was
necessary. Five of the ninety-four MBTI questions asked the respondent to choose between
three rather than the standard two responses. These questions were re-coded into
dichotomous variables.
The MBTI questions from each questionnaire were then subjected to a Principal Components
factor analysis, specifying a four-factor solution with Varimax rotation and pairwise deletion
of missing values. For Form A, the order of the factor loading for the scales was SN, JP, EI,
and TF, with four of the 94 items not loading in conformance with the theory (three from the
JP scale and one from the TF scale). For Form B, the order of the factor loading for the scales
was SN, TF, EI, and JP, again with four items not loading in conformance with the theory
(two from the JP scale, one from the TF scale and one from the IE scale). To determine
whether the reliability of the MBTI scales would be affected by the incorrect loading of these
items, reliability analysis was performed for each of the scales, with reverse coding of the
19
items that had loaded negatively on their respective factor. The results of this analysis are
shown in Table 7.
The reliability for all four scales in both forms of the research instrument was very good, with
coefficient alphas ranging from 0.76 to 0.88, and evincing little variation between the two
forms. Removing items that did not load correctly did not lead to a significant improvement
in the coefficient alphas. These reliability scores are consistent with those generally reported
in the literature (Gardner and Martinko, 1996:48). It can be concluded, therefore, that the
order of presentation of the MBTI and managerial style items had no significant effect on the
MBTI results, and so the cases from both forms of the research instrument can be combined
The factor and reliability analyses were repeated for the total sample. The order of the factor
loading for the scales was SN, JP, TF, and EI, with three of the 94 items loading incorrectly
(two from the JP scale and one from the TF scale). Table 7 also shows that the reliability
scores for the total sample are excellent, being in the range 0.80 to 0.86. Again, no increases
in reliability could be achieved by removing items from the scales. The reliability of MBTI
The literature generally indicates that the four MBTI scales are relatively uncorrelated, except
for the SN and JP scales, where coefficients in the order of 0.4 - 0.5 are common (Gardner
and Martinko, 1996:48). A correlation analysis was undertaken to determine the extent of the
correlation between the four continuous MBTI scales (cf Myers and McCaulley, 1985:9). The
20
As expected there is a significant correlation between the SN and JP scales (r = 0.51, p <
0.01). There are also significant correlations between the other scales, although these are
small and not unusual (cf Thomas, 1984:568; Myers and McCaulley, 1985: 151-153), the size
of the correlations depending to a large degree on the distribution of types in the sample. As
Myers and McCaulley (1985:154) point out, “larger than expected scale intercorrelations can
Table 9 shows, this is definitely the case for the current sample of managers, where the type
Normalisation of the type distribution, by creating weightings for the different types, was not
attempted. The US and UK type distribution data presented earlier shows that populations do
not fall equally into the 16 types, and so using weightings to create equality between the types
would be inappropriate. Further, there is no reported distribution of type for the Australian
population and so the appropriate weightings for an Australian sample are not known. In any
event, the use of weightings is not standard practice for MBTI or other personality descriptor
research. The type distribution for the total sample is shown in Table 9.
(52% to 48%), more intuitive (61%) and judging (67%), and overwhelmingly thinking (88%).
Chi-square analysis was undertaken to determine the extent to which the distribution in Table
9 was non-random. Without a population-based type distribution from which to derive the
expected cell counts, it was assumed for the purpose of the analysis that the expected cell
count for each cell was 32.62 (the sample size, n = 522, divided by the number of cells, 16).
21
As expected from inspection of Table 9, chi-square analysis showed that the type distribution
is clearly non-random (χ2df = 9 = 518.21; p < 0.000). In particular, E/ISTJs and E/INTJ/Ps are
over-represented, while there are fewer E/INFJ/Ps and E/ISFJ/Ps than expected. The most
obvious result from Table 10 is that relating to the TF distribution in the respondent group.
There are also sex differences in MBTI type distribution, at least as reported for the most
recent US national samples (Myers, et al, 1998:157:158), such that, on the TF dichotomy, a
significantly greater proportion of females (76%) than males (44%) are Fs, while a
significantly greater proportion of males (56%) than females (24%) are Ts. Table 10 presents
It was not possible to carry out a chi-square analysis of the distribution of type by sex because
16 cells (47.1%) had an expected count of less than five. Nonetheless, visual inspection of
the data indicates that ISTJs constitute a much larger percentage of the male sample (20%)
than for the female sample (5%). At the same time, the female group (20%) has a larger
percentage of INTJs than the male group (13%). These differences are reflected in the SN
dimension, where the SN split for males is approximately 60:40, while for females it is 70:30.
It might also be noticed that 52.3% of the current sample are NTs, constituted by 61.3% of the
As can be seen from Tables 9 and 10, the main types are E/ISTJ, E/INTJ, and E/INTP. The
most obvious feature of the type distribution, whether for the sample as a whole or by sex, is
the preponderance of Ts and Js. In this regard, a comparison of interest is that presented in
22
Table 11, between the current sample and a sample of 228 respondents from the School in
Myers and McCaulley (1985:39-40) commented on the high percentage (62%) of “tough-
minded TJs” in the 1980 sample. This percentage is reflected almost unchanged (60%) in the
current sample. Indeed, the most striking characteristic of Table 11 is the extent to which the
two profiles are similar in almost every respect, even though they are separated by almost two
decades, and the latter sample is nearly double the size of the earlier one.
DISCUSSION
The MBTI has been widely used in Australia for personality assessment, particularly as a
selection and development tool for managers. In the absence of appropriate, published local
norms, Boyle (1995) has cautioned against its routine use, especially in organizational and
occupational settings. The evidence presented here suggests his concern is well-founded, for
First, comparison of the results presented by Myers and McCaulley (1985) and Myers, et al
(1998) clearly indicates that the distribution of types differs between countries (cf Table 2).
While these differences are not large in raw percentage terms, it does mean, for example, that
there are twice as many INTJs in the US compared to the UK, while there are 50% more
ENFJs and ENTJs in the UK, compared to the US. This alerts us to the fact that it is unlikely
that the distribution of type in Australia will simply match that of either the US or UK and
Second, the results from Margerison and Lewis (1981) and Myers, et al (1998) show that, for
the UK at least, the distribution of type amongst managers differs markedly from that for the
23
general population (cf Table 3). This means that, even if the norms for the general population
are available, they cannot be simply applied to a particular sub-sample, for example,
managers. Finally, the results of the current study show that the distribution of type for a
sample of Australian managers is not only different from the UK sample as a whole, there are
also sex differences within the Australian sample (cf Tables 3, 9 and 10). Further, it is likely
that these results are consistently different over time (cf Tables 3 and 11).
At the same time, the results of the current study provide evidence of the reliability and
validity of the MBTI when used with Australian managers. These results also constitute a
significant contribution to the construction of an appropriate set of norms for ongoing use in
an Australian organisational setting. As such, they overcome one of the key concerns
expressed by Boyle (1995). Notwithstanding this important advance, it is argued that, even
though we now have a more solid foundation on which to base the use of the MBTI with
Australian managers, there are two sets of issues that suggest its use should continue to be
circumscribed. The first relates to the limitations of the current study, while the second relates
to whether, in any event, the MBTI is an appropriate tool for managerial or wider employee
selection.
A key potential limitation of the present study is that it involved a postal questionnaire with a
20% response rate. While this response rate is quite respectable in terms of postal surveys in
general, it is low. It might be argued that, given the low response rate, this is a group that
differs systematically from the non-respondents. On the other hand, the type distribution for
the current sample was almost identical to that of the earlier School group (cf Table 11) and
that survey was administered in a classroom context. Further it has already been noted that
the sample was representative of the alumni group as a whole. Of more concern is the
24
We know that certain psychological types are drawn to management as a career (cf Tables 2
and 3). There is some evidence to suggest that the group may not be representative of even
this narrower subset. For example, while the proportion of NTs in the current sample (52%)
is consistent with that of the earlier MGSM sample (47%, cf Table 11), it is significantly
higher than that for other population subgroups. Table 2 showed that the proportion of NTs in
the US and UK general population is between 9-10%, while Table 3 showed the proportion of
NTs in a general group of managers at 23%. Myers and McCaulley (1985:47-48) report much
lower percentages for US mature-age college students, whether female (13%) or male (24%).
It is not surprising that a group of people interested in possibilities, theories and concepts
might find their way to a post-graduate management education program, but these findings do
suggest that the alumni group carry with them a set of preferences that may not be common to
Given the preponderance of TJs and the apparent increasing social desirability of these
with the MBTI (Boyle, 1995; Furnham, 1990a, 1990b, 1990c; Gardener and Martinko, 1996).
Most of the examinations of ‘faking’ have been concerned with the use of the MBTI in job
selection. In this study however, the respondents were involved in research rather than job
selection, and their responses were provided on a confidential basis. It is unlikely then,
although not impossible, that social desirability was a major issue. That the result was so
similar to the sample from two decades earlier, when the MBTI was much less well-known,
adds further confidence that social desirability was not a major problem.
The limitations highlighted above point to the importance of further research to determine the
extent to which the findings here are replicable with other Australian managers. This could be
carried out as a part of a wider administration of the MBTI to a random, representative sample
25
of the Australian population. On the other hand, given the evidence of the differences
between managerial and broader populations, the research with additional manager samples
could proceed without the wider study. This brings us to the second set of issues concerning
its restricted use – even if we can use it, should we use it?
Use of the MBTI as a selection tool presumes some degree of predictive or criterion validity.
While this study has leant support to the content and construct validity of the MBTI, it has not
examined links between MBTI scores and the work behaviour of the respondents. It is not
possible therefore, on the basis of these results, to comment on whether the respondents
would exhibit certain managerial behaviours given a particular MBTI type. Indeed, there is
no published evidence linking the MBTI and managerial behaviour, broadly defined.
Previous literature concerning the relationship between personality as measured by the Myers-
Briggs Type Indicator, and managerial behaviour, can be grouped into two broad categories.
The first includes those studies that have atomised type and behaviour, focussing on one of
the dimensions of psychological type, for example SN or TF, and a specific managerial
behaviour, eg decision-making (cf Gardner and Martinko, 1996). The other category
comprises those books and articles that have made prognostications about managerial
behaviour based on defining managerial behaviour in terms of type, and so producing a self-
fulfilling prophecy (see, for example, Barr and Barr, 1989; Benfari, 1991, 1995; Isachsen and
Berens, 1989; Margerison and Lewis, 1981). The common assumption of these authors is that
knowledge of MBTI type can be used to predict particular styles of managerial behaviour.
Even then the question arises as to whether there is one best ‘type’. Perhaps it might be more
26
fruitful in selection and development to apply the principle of equifinality (cf Morgan,
1997:41; see also von Bertalanffy, 1950; von Foerester and Zopf, 1962). In this way, we
might benefit from the diversity, rather than the attempted homogenisation of management
27
REFERENCES
Anon (1991) “MBAiling: have business schools lost touch with business?” The Economist
Handbook of Modern Personality Theory. New York: John Wiley and Sons, 269-292
Boyle, GJ (1995) Myers-Briggs Type Indicator (MBTI): Some Psychometric Limitations.
Service
Consulting Psychologists Press (1994) 1994 Catalogue. Palo Alto, California.
Cooper, SE and Miller, JA (1991) MBTI learning style-teaching style discongruencies.
28
Costa, PT and McCrae, RR (1985) The NEO Personality Inventory Manual. Odessa, Fla.:
Resources.
Costa, PT and McCrae, RR (1993) Ego development and trait models of personality.
297-334
Croom, WC, Wallace, JM and Scheurger (1989) Jungian types from Cattellian variables.
Sage.
Furnham, A (1990a) Can people accurately estimate their own personality test scores?
46(7), 827-848.
Garden, A (1991) Unresolved issues with the Myers-Briggs Type Indicator. Journal of
29
Gardner, WL and Martinko, MJ (1996) Using the Myers-Briggs Type Indicator to study
managers: A literature review and research agenda. Journal of Management, 22(1), 45-
83.
Guion, RM (1991) Personnel assessment, selection and placement. In MD Dunnette and LM
Hough (eds) Handbook of Industrial and Organizational Psychology (2nd Ed). New
Hill.
Howes, RJ and Carskadon, TG (1979) Test-retest reliabilities of the Myers-Briggs Type
published 1921).
Kiersey, D. and Bates, M. (1984) Please Understand Me (5th Ed). Del Mar, CA: Prometheus
Nemesis Books.
Lorr, M (1991) An empirical evaluation of the MBTI typology. Personality and Individual
30
McCaulley, MH (1977) Applications of the Myers-Briggs Type Indicator to Medicine and
America, 381-418.
McCrae, RR and Costa, PT (1989) Reinterpreting the Myers-Briggs Type Indicator from the
perspective of the five factor model of personality. Journal of Personality, 57, 17-40.
More, E (1999) Personal communication. Professor More was Deputy Director of the
Macquarie Graduate School of Management in the period 1991-1996 and has been
Testing Service.
Myers, IB (1980) Introduction to Type (3rd Ed). Palo Alto, CA: Consulting Psychologists
Press.
Myers, IB and McCaulley, MH (1985) Manual: A Guide to the Development and Use of the
the Development and Use of the Myers-Briggs Type Indicator (3rd Ed). Palo Alto,
31
Pittenger, DJ (1993) The utility of the Myers-Briggs Type Indicator, Review of Educational
32
Thomas, AB (1993) Controversies in Management. London: Routledge.
Thomas, CR (1984) Regression of Myers-Briggs Type Indicator scales. Psychological
Australian business firms. Asia Pacific Human Resource Management, 27(4), 20-33.
von Bertalanffy, L (1950) The theory of open systems in physics and biology. Science, 3, 23-
29.
Von Foerester, H and Zopf, GW (eds) (1962) Principles of Self-Organization. New York:
Pergamon.
Wakefield, J, Sasek, T, Brubaker, M & Friedman, A (1976) Validity study of Eysenck
33
Table 1: Description of the Eight MBTI Preferences (adapted from Margerison and
Lewis, 1981:14)
34
Table 2: National Distributions of Type in the US and UK, rounded to the nearest
whole percent (adapted from Myers, et al, 1998:379)
35
Table 3: Type distribution for Margerison and Lewis’s (1981:16) manager sample
and the national distribution of type in the UK (adapted from Myers, et al,
1998:379), rounded to the nearest whole percent
36
Table 4: Summary of Demographic Data for the Respondents to the Main Survey
Questionnaire
37
Table 5: Pearson Correlations of Main Survey Sample Demographic and
Organizational Variables
1 2 3 4 5
1. Sex –
2. Age Groups .13** –
3. Place of Birth -.04 -.03 –
4. Management Qualification .10** -.08 .08 –
5. Level in Organisation .05 .17** -.08 .09* –
** Correlation is significant at 0.01 level (2-tailed). * Correlation is significant at 0.05 level (2-tailed).
38
Table 6: Chi-square analysis of individual and organizational variables
39
Table 7: Coefficient Alpha Reliability Scores of the MBTI Scales for the two forms
of the research instrument, separately and combined
40
Table 8: Pearson Product Moment Correlations of the MBTI Continuous Scales
41
Table 9: MBTI Type Distribution for the Total Sample (n= 523)
42
Table 10: MBTI Type Distribution for the Total Sample by Sex (Female n= 118;
Male n = 394)
Female Male
Sensing Types Intuitive Types Sensing Types Intuitive Types
with With with with with with with with
Thinkin Feeling Feeling Thinkin Thinkin Feeling Feeling Thinkin
g g g g
ISTJ ISFJ INFJ INTJ ISTJ ISFJ INFJ INTJ
n=6 n=2 n=2 n = 24 n = 79 n=4 n=6 n = 50
% = 5.0 % = 1.7 % = 1.7 % = 20.2 % = 20.1 % = 1.0 % = 1.5 % = 12.7
ISTP ISFP INFP INTP ISTP ISFP INFP INTP
n=1 n=1 n=2 n = 11 n=6 n=4 n=8 n = 40
% = 0.8 % = 0.8 % = 1.7 % = 9.2 % = 1.5 % = 1.0 % = 2.0 % = 10.2
ESTP ESFP ENFP ENTP ESTP ESFP ENFP ENTP
n=0 n=2 n=5 n = 20 n=8 n=1 n = 12 n = 46
%=0 % = 1.7 % = 4.2 % = 16.8 % = 2.0 % = 0.3 % = 3.1 % = 11.7
ESTJ ESFJ ENFJ ENTJ ESTJ ESFJ ENFJ ENTJ
n = 18 n=1 n=5 n = 18 n = 60 n=5 n=5 n = 59
% = 15.1 % = 0.8 % = 1.3 % = 15.1 % = 15.3 % = 1.3 % = 1.3 % = 15.0
E 69 (58%) > I 49 (42%) E 196 (50%) = I 197 (50%)
S 31 (29%) < N 87 (71%) S 167 (42%) < N 226 (58%)
T 98 (91%) > F 20 (9%) T 348 (89%) > F 45 (11%)
J 76 (70%) > P 42 (30%) J 268 (68%) > P 125 (32%)
43
Table 11: Comparison of 16 MBTI Types for two samples from the School,
separated by two decades. 1980 data from Myers and McCaulley,
(1985:39-40)
44