Vous êtes sur la page 1sur 15

Journal of Occupational and Organizational Psychology (1992), 65, 159-172 Printed in Great Britain

(E) 1992 The British Psychological Society


The personality of UK managers: 16PF norms
for short-listed applicants
Dave Bartram*
Department of Psychology, University of Hull, Hull HU6 7RX
There is little information available on the performance of UK managers on the 16PF
when administered as part of a selection procedure. The present paper reports the results
of analyses of data from 1796 short-listed managers who were given the 16PE (Eorm A).
The assessmenr was part of a selection procedure carried out for a wide range of clients
by an executive recruitment agency. Normative analyses of these data are presented and
effects related to age and gender are examined. Internal consistencies of both primary
and secondary scales are presented and the problem of low primary reliabilities is dis-
cussed. Eactor analysis of the primary scale scores produced five second-order factors
which closely match the 16PE second-order factor scales.
Despite the widespread use of the 16PF in management selection in the UK, no large-
scale British study has been reported which examines the performance of managers on the
16PF in a selection situation. The majority of users rely on the British General Population
Standardisation (Saville, 1972) and the British Undergraduate Norms (Saville &
Blinkhorn, 1976). While this may pose no practical problem when the 16PF is being used
for individual interpretative development purposes, it does raise problems in selection.
Use of general population norms tends to carry with it an implicit referencing of appli-
cants against those norms. Thus, if a company are looking for an 'extravert manager', they
will tend to perceive scores on Extraversion which are in the region of 7 upwards as indi-
cating an extraverted person. However, it may be that most if not all managers in a selec-
tion situation tend to score in that region (either because, as a group, they tend to be
relatively extravert or they are 'dissimulating' or both). By comparing them against norms
established for other managers assessed under the same conditions one would get a better
picture of their 'relative' extraversion.
Given the increasing use of inventories like the l6PF in management selection, it is
particularly important to know how applicants for management positions typically
respond in a selection situation. The population who come forward for Selectionmay well
be unrepresentative of the general population (through self-selection) and scores obtained
during a selection process may be subject to a degree of distortion from impression
management. Comprehensive norms developed from people undergoing selection should
enable effects of self-selection and possible impression management to be taken into-
account when interpreting applicant profiles. Individual companies and consultancies
* Requests for reprints.
160 Dave Bartram
who use 16PF tend to create and use their own 'local' norm databases. While some of these
may be quite large, access to them tends to be restricted.
As a 'reference' population, managers undergoing selection may differ from the general
popuiarion not only in their average scores on each scale, but also in the spread of scores
obtained. Such range restrictionif it occurswould reduce scale reliability and hence
the ability of the inventory to discriminate between people.
This paper presents the results of analyses carried out on a large national sample of UK
managers using data collected with the SCREENTEST computer-based version of 16PF
(Form A). It presents information on the reliability of both first- and second-order scale
scores and on age-related and gender-related differences.
Method
The 16PF
The SCREENTEST version of the Sixteen Personality Factor Questionnaire (16PF) Form
A was used (see Bartram, 1989a,b). This is a fully automated version of 16PF (i.e. no
paper-and-pencil materials are used). The SCREENTEST 16PF items are those from the
anglicized 1978 versions as published by ASE. Previous research (e.g. Harrell &
Lombardo, 1984) has suggested that automation of the 16PF has no significant effect on
scale score retest reliabilities or validiries.
In the SCREENTEST version, items appear on the screen one at a time. There is an
option to return to the immediately preceding item to correct inadvertent keying errors.
However, the candidate is not free to 'browse' through the items. On average, the time
taken to complete the SCREENTEST 16PF was about 50 per cent less than for the paper-
and-pencil version.
SCREENTEST saves both raw scale scores for each test and item responses. A small
number of paper-and-pencil test administrations were added to the SCREENTEST
database. Where this was done, only the raw scores were entered. So, item data were avail-
able for all the SCREENTEST administered sessions, but not for the paper-and-pencil
ones.
The sample
The sample comprised 1796 people (1441 male and 355 female), 1732 who completed
the SCREENTEST version and 64 who completed the paper-and-pencil version. They
were all short-listed candidates for managerial appointments who had some previous
management experience (i.e. they were not graduate applicants for management training
schemes). Data were collected over the period 1988-90 at a number of sites around the
UK: London, Windsor, Bristol, Birmingham, Manchester and Leeds. Most of the data
were collected in Manchester and London. Candidates were applicants for a wide range of
both public and private sector managemenr positions. In most instances, assessments
were carried out on short-listed applicants for parricular posts (typically half a dozen or so
people per post). Given the fact that candidates were sampled from across the whole coun-
try for a very wide range of industries, the final sample should be reasonably representa-
tive of the population of UK managers and supervisors.
The personality of UK managers
Table 1. Breakdown of the Management Sample by age and gendet
161
Age
Less than 25
25-34
35-44
45-54
55 or over
Total
Gender
Female
46
134
60
20
0
260
Male
32
452
486
220
14
1204
Total
78
586
546
240
14
1464
Pet cent female
59
23
11
8
0
18
Infotmation about age was obtained fot 1464 of the candidates (1204 males and 260
females: see Table 1). The percentage of females reduces as age increases. Females are in
the majority in the under 25 age group, represent about a quarter for the 2534 age group
and only about 10 per cent of the older groups.
Results
Differences between the average profile and the general population
The average sten-score profile, using UK general population (male and female) nornas,
is shown in Fig. 1. Second.-order factor scores (computed using the equations given in
10
STEN Scores
10
1
E F G H I L M N O Q1 Q2 Q3 Q4
First order factor scales
All n-1796 Females n-355 Males n-1441
Figure 1. Average r6PF first-order scale sten scores for Management Sample (stens based on UK general
population combined male and female norms).
162 Dave Bartram
STEN score
9-
8-
/
6-
5-
4 -
3 -
2 -
1 - ;: HI ;
^ ^
1 m
Extraversion Anxiety Tough Poise independence
Second order factors
Control
I Aii n-1796 Maie n-1441 Femaie n-355
Figure 2. Average 16PE second-order scale sten scores for Management Sample (stens based on UK general
population combined male and female norms).
Table 2. Raw score means and SDs for the Management Sample and the UK
general population (the latter are from Saville, 1972)
Variable
16PF-A
A
B
C
E
F
G
H
I
L
M
N
O
Qi
Q2
Q3
Q4
General
population
(2220)
M
10.22
7.01
14.64
10.86
12.36
12.26
12.47
11.25
8.25
12.02
11.16
11.44
9.11
11.50
12.36
13.16
SD
3.06
2.13
3.90
4.38
4.99
3.56
5.46
3.75
3.44
3.55
3.18
4.30
3.31
3.51
3.23
4.92
All (1796)
M
11.85
9.43
17.88
15.98
16.88
14.09
19.37
9.38
6.84
14.69
7.89
6.78
11.31
8.99
13.80
8.16
SD
3.30
1.88
3.70
3.87
4.20
3.42
4.93
4.06
3.24
3.30
2.97
3.81
3.25
3.42
3.04
4.86
Management Sample
Male(
Af
11.79
9.54
18.03
16.24
16.79
14.26
19.62
8.58
6.90
14.83
7.75
6.29
11.61
8.96
13.94
7.71
1441)
SD
3.36
1.82
3.67
3.75
4.14
3.36
4.78
3.79
3.23
3.26
2.86
3.56
3.13
3.36
2.96
4.59
Female
M
12.08
8.95
17.27
14.90
17.22
13.39
18.37
12.66
6.60
14.11
8.46
8.78
10.08
9.14
13.24
9.96
(355)
SD
3.02
2.06
3.78
4.14
4.40
3.57
5.37
3.41
3.29
3.43
3.34
4.13
3.41
3.64
3.29
5.48
The personality of UKmanagers 163
Krug & Johns, 1986) are shown in Fig. 2. Raw score means and SDs for each scale, with
the UK general population means for comparison, are shown in Table 2.
By definition, a profile of the average sten scores for the general population would be a
horizontal line on Fig. 1 at 5.5. Differences between the mean general population raw
scores and the Management Sample's mean raw scores are statistically significant for sill
16 scales (absolute t (4014) = > 13 in all cases, ^ < .001). Examination of Figs. 1 and 2
shows that most of these differences are substantial (that is, the Management Sample aver-
age differs from the general population average by more than one sten). Both males and
iSsmales show a similar partern of differences from the general population norm.
Gender-related difference
Given the sample size, even quite small differences between groups are statistically
significant. Only four scales did not show gender-related differences (A, F, L and Q2: see
Table 3). For the other 12 scales, the differences berween the means for males and females
were statistically significant (absolute t (1794) = > 3.48,^ < .001 in all cases). For the
second-order factors only Exrraversion did not show a gender-related difference. In the
other four cases, the means for males and females were significantly different (absolute t
(1794) = > 4. 63, / ' < .001 in all cases).
However, many of the observed gender-related differences are quite small (half a sten
score or less) and hence are of little substantive importance. For three scales, however,
the differences between males and females are both significant and substantive: I (tough-
mindedness), O (self-assuredapprehensive) and Q4 (tenserelaxed). The various first-
Table 3. Differences between raw score means for males and females for the general
population sample and the Management Sample: 16PF form A
Variable
16PF-A
A
B
C
E
E
G
H
I
L
M
N
O
Qi
Q2
Q3
Q4
Male
(N = 1441)
11.79
9.54
18.03
16.24
16.79
14.26
19.62
8.58
6.90
14.83
7.75
6.29
11.61
8.96
13.94
7.71
Management Sample
Eemale
(N = 355)
12.08
8.95
17.27
14.90
17.22
13.39
18.37
12.66
6.60
14.11
8.46
8.78
1O'.O8
9.14
13.24
9.96
Diff
Male-Eem
- 0 . 2 9
0.59
0.76
1.34
- 0. 43
0.87
1.25
- 4. 08
0.30
0.72
-0. 71
-2. 49
1.53
- 0. 18
0.70
-2. 25
lvalue
(d. = 1790)
1.47
-5. 32
-3. 48
-5. 94
1.74
- 4. 33
-4. 29
18.51
-1. 56
-3. 69
4.05
11.42
- 8. 08
.87
- 3. 94
7.95
Per cent
Var
0.12
1.36
0.72
1.94
0.17
1.03
1.00
16.23
0.12
0.71
1.13
6.67
3.59
0.03
0.83
3.51
UK general population
Male
(N= 1104)
9.35
7.17
15.25
12.28
12.39
12.47
13.25
9.16
8.64
12.37
10.52
10.27
10.12
11.60
12.79
11.99
Female
fN=1113)
11.09
6.85
14.03
9.45
12.34
12.05
11.70
13.33
7.87
11.65
11.81
12.58
8.11
11.40
11.94
14.31
DifF
Male-Feni
-1. 74
0.32
1.22
2.83
0.05
0.42
1.55
-4. 17
0.77
0.72
-1. 29
-2. 31
2.01
0.20
0.85
-2. 32
Note. Effect sizes for the Management Sample are shown as raw score differences (Diff Male-Fem) and as the percentage
of scale score variance accounted for by gender (per cent Var).
164 Dave Bartram
order differences combine to produce a clearer pattern of difference between males and
females for tbe second-order factors. These show clear differences of about one sten for
Anxiety, Tough-Poise and Independence (with males scoring on average lower on Anxiety
and bigber on Tough-Poise and Independence). Tbere is also a difference of over half a sten
on Control (witb the male average being lower than the female one).
Age-related differences
The average age ofthe 16PF sample was 36. Tbe female group was younger than the male
group (31.62 as opposed to 37.12). A number of scales (see Table 4) sbow small but
statistically signii&cant correlations with age. Older candidates tend to be more 'Sober'
(F), 'Conscientious' (G+), 'Imaginative' (M+), 'Self-sufficient' (Q2-I-) and 'Socially
precise' (Q3+). Table 4 also shows that there are some differences between tbe age-related
effects for the male and female groups.
Table 4. l6PF (Form A) correlations with age
Scale
A
B
C
E
F
G
H
I
L
M
N
O
Qi
Q2
Q3
Q4
Age
*p<
A11(N
At
11.88
9.42
18.08
16.04
16.89
14.23
19.66
9.27
6.75
14.70
7.81
6.59
11.39
8.93
13.97
7.84
36.14
05; **p < .01
= 1311)
SD
3.33
1.89
3.64
3.84
4.20
3.35
4.75
4.10
3.18
3.27
2.98
3.78
3.23
3.47
3.00
4.71
7.92
r
.03
.06
.03
.03
-. 28**
.14**
.03
-. 09**
-. 02
.15**
.04
- . 15**
.07*
.10**
.11**
- . 08*
Females
M
12.10
8.74
17.51
14.73
17.27
13.48
18.66
12.62
6.62
14.06
8.43
8.97
10.05
8.90
13.52
9.67
31.62
(N =
SD
3.03
2.06
3.74
4.23
4.46
3.50
5.08
3.57
3.22
3.19
3.35
4.17
3.38
3.66
3.33
5.31
7.46
234)
r
- . 04
.19*
.10
.05
29**
. 21**
.04
-. 05
-. 20**
.20**
- . 10
-. 17*
.12
.12
.13
- . 14
Males (N =
M
11.83
9.56
18.20
16.32
16.80
14.40
19.87
8.55
6.77
14.84
7.67
6.07
11.68
8.93
14.07
7.44
37.12
SD
3.39
1.82
3.61
3.69
4.14
3.30
4.64
3.83
3.17
3.27
2.85
3.49
3.12
3.43
2.92
4.48
7.68
1077)
r
.06
-. 02
- . 01
- . 03
_ 27**
.10**
- . 09
.03
.01
.12**
-. 06
- . 01
.10**
.09*
- . 00
Wbile age- and gender-related effects on each scale do not interact statistically, tbe dis-
tributions of people by age and gender are somewhat confounded in this sample (in that
tbe proportion of females reduces as age increases, and tbe overall average age for females
is lower than for males). Nevertheless, it does seem that tbe 'age' effects seen in Table 4
are different in kind from the 'gender' effects seen in Table 3. For the most part different
scales are involved in each. The large gender-related differences are seen on scales E, H, I,
The personality of UK managers 165
O, QI and Q4, while the main age-related effects are on scales F, G, M, O, Q2 and Q3.
Smaller age-related effects are seen for scales I, QI and Q4. Thus, it is likely that some of
the age related variance for I, O, QI and Q4 may be due to confounding effects of gender
difference. It is unlikely to be the other way round (i.e. apparent gender diflferences aris-
ing because of age effects) as the gender differences on these scales are well-established for
general population groups of matched ages.
Reliability analysis
Complete item data for Form A were available for 1732 people in the sample. The itenci
responses were converted to item scores (as specified by the scoring key) and then sorted
into 16 sets according to their respective scales. The item scores for each set of scales were
then analysed using the SPSS/PC + RELIABILITY procedure. Alpha coefficients for each
scale with and without adjustments for range restriction are shown in Table 5. Two scales
(A and I) had larger raw score SDs for rhe present sample than for the general populatioiti
sample. In all other cases there was some degree of range restriction.
Table 5. 16PF Form A: reliabiliry estimates
Scale
A
B
C
E
IP
G
H
I
I
M
N
O
QI
Q2
Q3
Q4
N =
M
12.00
9.34
18.01
16.15
16.98
14.26
19.67
9.35
6.80
14.74
7.76
6.70
11.45
8.92
13.85
8.13
Management Sample
1732
SD
3.33
1.91
3.71
3.82
4.18
3.43
4.82
4.12
3.21
3.28
2.98
3.79
3.21
3.45
2.01
4.88
Alpha
(a)
.42
.47
.54
.51
.62
.53
.77
.56
.40
.18
.26
.57
.37
.44
.40
.70
(b)
.31
.58
.59
.61
.74
.56
.82
.47
.47
.30
.35
.67
.41
.A6
.48
.71
SEm
(b)
1.66
1.30
1.28
1.23
1.02
1.33
0.85
1.46
1.46
1.67
1.61
1.15
1.54
1.47
1.44
1.08
UK general
population
AvB
(c)
0.51
0.44
0.51
0.57
0.65
0.47
0.73
0.56
0.31
0.27
0.25
0.61
0.25
0.41
0.51
0.66
SEm
(c)
1.40
1.50
1.40
1.31
1.18
1.45
1.04
1.33
1.66
1.71
1.73
1.26
1.73
1.53
1.41
1.17
(a) Uncortected standardized item alpha.
(b) Adjusted for range restriction in comparison with genetal population norms.
(c) Correlations between alternate Forms A and B (N = 2007) (Saville & Munro, 1986).
Of the 16 scales, only three had corrected alphas greater than .70 (F, H and Q4). Two
had alphas between .60 and .69 (E and O) and three had alphas between .50 and .59 (B,
166 Dave Bartram
C and G). The remaining eight had alphas less than .50. Across the scales, there is a clear
relationship between the magnitude of alpha and that of the alternate form correlation.
Alpha is higher than the alternate form reliability for 13 of the 16 scales.
Reliability of the second-order factors
Internal consistency (Cronbach's alpha) for second-order factor scores were computed
using the procedure described by Mosier (1943) from: the primary scale intercorrelation
matrix; the primary factor alphas (corrected for range restriction); and the Krug & Johns
(1986) second-order factor score weights. These internal consistencies (see Table 6) are
good to high, producing standard errors of measurement which are less than one sten (i.e.
half an SD). These figures are comparable with the short-term and longer-term retest
correlations quoted in the 16PF Administrator's Manual (1986). For the second-order
factor scores one can be reasonably confident that a person's true score is within one sten
of their obtained score.
Table 6. 16PF (Form A) second-order factors
Variable
Extraversion
Anxiety
Tough-Poise
Independence
Control
M
1.1'b
3.09
5.36
8.00
6.61
SD
1.82
1.83
1.70
1.82
2.02
Alpha
.90
.92
.71
.89
.82
SEm
0.58
0.52
0.92
0.60
0.86
Note. Scores obtained by applying 1986 second-order facror score equations to sten
scores obtained using UK general population (combined male and female) norms
for Form A.
Robustness of the second-order factor structure
While Cattell, Eber & Tatsuoka (1970) describe eight 'second stratum' factors, common
practice is to use only five of these (QI, QII, QIII, QIV and QVIII) as QV and QVI are not
well-defined and QVII corresponds directly to the first-order Scale B. Robustness was
assessed by generating factor scores from the present data and then correlating them with
the five second-order factor scale scores produced using the Krug & Johns (1986) equa-
tions.
The raw score scale intercorrelation matrix was computed with Factor B omitted.
Examination of the principal component eigenvalues (Scree test) suggested five factors
should be extracted (the five components accounted for 63.5 per cent of the total vari-
ance). Principal axis factoring with Direct Oblimin rotation was used to produce loadings
and factor score coefficients for the five factors (see Table 7). Hakstian (1971) found this
the best procedure for rotating to 'simple structure' (Cattell, 1978). Factor scores were
produced for each of the rotated factors using the regression method. These scores were
then correlated with the original second-order factor sten scores. No attempt was made to
factor separately the male and female data.
The personality of UK managers 167
Table 7. Direct Oblimin rotated pattern matrix of factor
loadings
16PF ^^S!25
scale I II III IV V
Q4
C
O
L
E
Ql
N
F
H
Q2
A
G
Q3
M
I
- . 78
.66
- . 64
-. 47
-. 05
.08
- . 04
- . 01
.26
- . 04
.06
-. 12
.21
.15
- . 21
- . 02
.02
- . 21
.38
.70
.53
-. 39
.17
.26
.08
.03
.04
- . 14
.22
- . 21
- . 12
.15
-. 02
.02
.18
-. 02
- . 19
.63
.60
-. 57
.49
.04
.04
- . 14
.17
- . 08
-. 02
- . 09
- . 03
- . 06
- . 03
.10
-. 16
.06
-. 07
.09
.79
.62
-. 02
- . 09
-. 02
- . 03
- . 04
-. 05
- . 03
.07
-. 06
- . 08
.03
.06
.10
- . 03
.02
.51
.37
Table 8. Correlations between second-order factor sten scores and Management
Sample factor scores MANl to MAN5 (N = 1796)
Extraversion
Anxiety
Tough-Poise
Independence
Control
MAN3
MANl
MAN5
MAN2
MAN4
Extraversion
1.00
-. 42
-. 12
.46
.08
.97
.34
.09
.36
.07
Anxiety
1.00
- . 01
- . 26
-. 37
-. 39
- . 98
- . 11
- . 20
-. 47
Correlations between the five Management
MAN2
MAN3
MAN4
MAN5
MANl
.19
.30
.AA
.14
MAN2
.37
- . 29
.09
Tough-Poise
1.00
-. 14
.16
- . 14
.02
- . 88
- . 06
.22
Sample factor scores
MAN3 MAN4
.04
.09 - . 34
Independent
1.00
-. 27
.49
.23
.15
.95
- . 31
Control
1.00
.07
.33
- . 31
- . 24
.97
(MANl to MAN5)
168 Dave Bartram
Inspection of the pattern of loadings in Table 7 and the correlations in Table 8
shows a very clear correspondence of Factor I with Anxiety (r = .98), Factor II with
Independence (r = .95), Factor III witb Fxtraversion {r = .97), Factor IV with Control
(r = .97) and Factor V with Tough-Poise (r = .88). (Note that the direction of the
Anxiety and Tough-Poise scales are reflected.) Tbese results provide strong support for
the robustness of tbe five factors originally derived from the United States general popu-
lation data.
Discussion
The 'management' profile
The profile of the 'average' management position applicant can be characterized as:
'An independent, stable extravert who is neither particularly tough-minded nor tender-
minded but who is somewhat more controlled and conventional tban the average for the
general population'.
As these are, by and large, the sort of qualities one would expect to see in managers,
this profile may reflect a degree of 'impression management' on the part of candidates.
However, it may also be quite 'genuine': people with a background in management wbo
apply for management positions are more likely to possess the qualities needed for (or
expected in) such a position. In practice, of course, it is difficult to detect impression man-
agement simply on the basis of test scores. All candidates were given a 16PF feedback
interview as part ofthe assessment process. In addition, the Karson & O'Dell (1976) 'fak-
ing good', 'faking bad' and 'random responding' scales were always examined. To the
extent that differential effects of'faking good' are detectable in feedback interviews, these
effects would have been 'controlled for ' in the qualitative statements made by consultants
in their final reports.
There bas been much recent debate about the degree to which 'impression manage-
ment' or 'faking good' affects scores on personality inventories (e.g. Edwards, 1990;
Nicholson & Hogan, 1990; Walsh 1990). Paulhus (1986) argues that two factors are
involved in effects measured by 'social desirability' scales: self-deception and 'impression
management'. The latter involves a conscious attempt to make oneself'look good' and is
the cornponent most likely to affect people's scores on inventories when they are taken in
a selection situation.
It is well established that people can 'fake' personality inventories under laboratory con-
ditions (e.g. Power & MacRae, 1977). Whether they actually do so under real-life condi-
tionswhere they will be subject to checks from other assessment proceduresis less
clear. As Leary & Kowalski (1990) argue, people have both to be able to create a particu-
lar impression and; be motivated to do so. To provide a direct measure of the effects of the
selection process on 16PF scores, it would be necessary to look at retests for the same
people (or a carefully matched group) assessed botb at selection and after selection.
Tbe fact that the deviations from the general population norm might be partially or
even wholly attributable to the effects of positive impression management does not
necessarily reduce tbe utility ofthe 16PF as a source of information for selection purposes.
The variance on each scale for this sample is very little reduced from that found for the
general population, hence variations within the group can still be examined. Indeed, the
The personality of UK managers 169
present norms can be used in place of the population norms to profile individual variations
around the management average. This provides a degree of control for general impression
management effects but does not allow for the possibility of some people showing more
or less distortion than others.
Use of sten based on these management norms will also provide better discrimination
between candidates than stens based on the general population norms (normalized sten-
scote conversion tables have been produced from the Management Sample data for the
SCREENTEST 16PF Profiler). As the distributions of such scores are normalized around
the relevant sample mean rather than the general population mean, 'floor' and 'ceiling'
effects will be reduced. However, one has to be careful in the sort of interpretive state-
ments one uses about people if the stens are based on norms for a group which is very dii-
ferent from the general population.
Gender-related and age-related differences
In general, gender-related differences found in the Management Sample reflect the
general pattern of gender differences one finds in the UK general population
(Saville, 1972), though the differences between the mean male and female raw scores
tend to be somewhat smaller for the present sample than for the general pop-
ulation. This suggests that the general population differences are 'overlaid' on a manage-
ment group profile with only a slight reduction being caused by differential selection
effects.
The issue of whether separate or combined male and female norms should be used is a
complex one. In practice, without evidence for or against differential job-related validity,
the overriding principle niust be to interpret scores with due regard for the nature of the
norm groups used to derive them and the purpose of the interpretation. In the present
case, the female group can be considered to have been representative of shont-listed
women applicants for the same range of management posts as the male candidates. The
data also showed that they differ, as a group, from males in much the same manner as the
general population male and female groups differ from each other.
The possible confounding effects of age (the female group had a lower mean age than
the male group on the present sample) may have slightly exaggerated differences on some
of the scales. However, age-related effects were different in kind from gender-related
effects.'
The changing ratio with age of males to females raises a number of questions. It could
reflect the fact that more women are now going into managencient than used to, and that
there is therefore differential availability across the age bands. In this case, one would
expect the male:female ratio to gradually shift towards 50:50 as one moves up the age
range as rhis effect works its way through over the next 20 years or so. On the other hand,
it may be that the ratio of males to females is distorted by an increasing bias towards men
for more senior (and hence older) positions. The present data do not allow us to distin-
guish between these two alternativesnor are rhey mutually exclusive. As the pool of
trained women who are qualified in management increases in size and experience, so we
should isee an increasing number of applicants for more senior positions and a concomi-
tant increase in the number of appointments. To the extent that increases in appointments
170 Dave Bar tram
do not keep pace with increasing experience, qualification and availability for appoint-
ment, there is a problem of unfair discrimination.
Reliability issues
There is considerable debate about how best to estimate reliability for personality inven-
tory scales. Cattell et al. (1970) argue that items are chosen to form a single scale because
they correlate significantly with the underlying factor^they do not necessarily have to
correlate significantly with each other. Indeed Cattell has argued that it is inappropriate
to use scale homogeneity (e.g. as measured by coefficient alpha or KR-20) as a measure of
reliability. Catell prefers to use retest measures to estimate reliability: short-term same-
form retest giving a measure of dependability, with long-term same-form retest provid-
ing information about trait stability. (For a more detailed discussion of Cattell's approach
to reliability issues the reader is referred to chapter 5 of the 16PF Handbook: Cattell et al.,
1970.)
Others (e.g. Saville & Munro, 1986) have argued that while Cattell may be right in
theory that 'breadth' can only be obtained at the expense of low internal scale consistency,
in practice such scales are also unreliable in terms of alternate form correlations or retest
measures. While there were no retest data available for the present sample, it was felt
worthwhile to compare measures of internal consistency for the 16 scales with alternate
form and retest reliability esrimates from other sources.
The 16PF Handbook (Cattell et al., 1970) and the more recent Admiyiistratnr's Manual
(1986) give a number of short-term and longer-term retest reliabilities. In addition, they
provide a range of equivalence coefficients. Examination of these indicate that the esti-
mates obtained from the Management Sample are generally consistent with those c]uoted
for Form A vs. B equivalence and for the longer-term retest measures. For short-term
retests (immediate to two weeks), much higher 'dependabilities' are quoted. Three short-
term retest samples (containing 243, 146 and ^\ people respectively) are quoted by
Cattell el al. (1970). The lowest retest correlation is .58 for Scale B. All others arc above
.60 with half being .80 or greater. Thus, short-term retest data (though somewhat lim-
ited in extent) does indicate reasonable 'dependability' for ail scales.
Cattell repeatedly stresses the point that one should not expect to obtain high 'relia-
bility' with short scales and also expect good construct validity (i.e. breadth of coverage).
He strongly advocates use of two forms rather than one (Forms A and B), though this rec-
ommendation is rarely followed as the administration rime required is usually felt to be
excessive. The fact that computer administration reduces the time taken for candidates to
complete the 16PF by about 30 per cent does make it more practical to increase reliability
by administering both forms.
For practical purposes, artention should focus on the standard error of measurement
(SEm) rather than reliability estimates. Using the average Form A short-term depend-
ability as an estimate of reliability, one would expect SEois to be around 0.89 sten scores.
Both the internal consistency and the ecjuivalence measures suggest that this may be opti-
mistic. If, as Cattell argues, internal consistency underestimates the effective reliability of
the measures, then the SEm values of around 1.5 stens computed from alternate form or
internal consLstency data will be conservative and provide the 'worst case' situation.
The personality of UK managers 171
Taking the less conservative sbort-term retest data to provide an estimate of reliability
reduces the SEm to around 1 sten.
However one interprets the evidence, it may be safest to regard the standard error of
measurement for most primary scales as being about 1.5 stens rather than under 1 sten.
The main exceptions are F, H, O and Q4, which all seem to have reasonable consistency,
reasonable alternate form and retest reliabilities, and can safely be regarded as having
SEms of around 1 sten.
The reliability of the first-order scales needs to be considered in relation to tbat of the
second-order ones: these were seen to be quite bigh. In practice, first-order scores are used
to aid the interpretation ofthe second-order ones. The former describe 'facets' of persotii-
ality within the broader domains covered by the latter.
Second-order factor structure
The factor analysis ofthe present data closely replicated the expected five-factor stmcture:
factor scores derived from this analysis correlated very highly with the commonly used
five second-order factor scales (.88 or better). The present data did not provide any evi-
dence for the additional two factors (QV and QVI) postulated by Cattell.
There is an increasing consensus on the identity and generality of a relatively small
number of domains of personality (e.g. Digman, 1990; McRae & Costa, 1987) and on
their utility as a framework for studies of validity (Barrick & Mount, 1991). Often
referred to as the 'Big Five' factors, tbese have been labelled as Extraversion;
Agreeableness/Independence; Conscientiousness; Anxiety and Openness to Experience.
While some of these (Extraversion and Anxiety) are better defined than others, there is a
clear overlap between them and the 16PF second-order factors.
Conclusions
The present data provide useful new normative and psychometric information on the
16PF when administered to short-listed candidates for managerial posts. Further research
will be needed to assess the degree to which the departure ofthe profile from the general
population norm represents effects of self-selection, preselection or impression manage-
ment. In practice, it is likely that all three of these factors contribute to the obtained
profile, and tbe major research question should now be bow much each contributes.
Acknowledgement
The author is grateful to Roy Hammond and other staff and consultants at MSI.
International for their help and assistance in collecting and assembling the data used in
this study.
References
Anon (1986). Administrator's Manual for the 16 Personality Factor Questionnaire. Champaign, IL: IPAT.
Barrick, M. R. & Mount, M. K. (1991). The Big Five personality dimensions and job performance: a meta.-
analysis. Personnel Psychology, 44, 1-26.
Barrram, D. (1989<). SCKEESsSTBSr Assessment and Data Manager User's Manual. Windsor: NFER-NELSON.
172 Dave Bartram
Battram, D. (1986*). Sixteen Personality Factor Questionnaire: Software Documentation for the SCREENTEST Test
Adminstration Modules. Windsor: NFER-NELSON.
Cattel, R. B. (1978). The Scientific Use of Eactor Analysis in Behavioral and Life Sciences. New York: Plenum Press.
Cattell, R. B., Eber, H. W. & Tatsuoka, M. M. (1970). Handbook for the Sixteen Personality Factor Questionnaire.
Champaign, IL: IPAT.
Digman, J. M. (1990). Personality structure: Emergence of the five-fector model. Annual Review of Psychology,
41, 417- 440.
Edwards, A. L. (1990). Construct validity and social desirability. American Psychologist, 45, 287289.
Hakstian, A. R. (1971). A comparative evaluation of several prominent methods of oblique factor transfor-
mation. Psychometrika, 36, 175-193.
Harrell, T. H. & Lombardo, T. A. (1984). Validation of an automated 16PF administration procedure .Jwrwa/
of Personality Assessment, 48, 638-642.
Karson, S. & O'Dell, J. W. (1976). A Guide to the Clinical Use of the 16PF. Champaign, IL: IPAX
Krug, S. E. & Johns, E. F. (1986). A large-scale cross-validation of second-order personality structure as deter-
mined by the 16PR Psychological Reports, 59, 683-693.
Leary, M. R. & Kowalski, R. M. (1990). Impression management: a literature review and two-component
model. Psychological Bulletin, 107, 34-47.
McRae, R. R. & Costa, P. T. (1987). Validation of the five-factor model of personality across instruments and
ohservets. Journal of Personality and Social Psychology, 52, 8190.
Mosier, C. I. (1943). On the reliability of a weighted composite. Psychometrika, 8, I 6I - I 68.
Nicholson, R. A. & Hogan, R. (1990). t h e construct validity of social desirability. American Psychologist, 45,
290-292.
Paulhus, t ) . L. (1986). Self-deception and impression management in test responses. In A. Angleitner &J. S.
Wiggins (Eds), Personality Assesssmem by Questionnaire. Berlin: Springer-Verlag.
Power, R. & McRae, K. (1977). Characteristics of items in the Bysenck Personality Inventory which affect
respotjses when students simulate. British Journal of Psychology, 6S, 491498.
Saville, P. (1972). The British Standardisation of the I6PF Supplement of Norms, Forms A andB. Windsor: NFER
Publishing.
Saville, P. & Blinkhorn S. (1976). British Undergraduate Norms to the 16PF (Forms A and B), Windsor: NFER-
NELSON.
Saville, P. & Munro, A. (1986). The relationship between the Factor Model of the Occupational Personality
Questionnaire and the 16PF. Personnel Review, 15, 30-34.
Walsh, J. A. (I99O). Comment on social desirability. American Psychologist, 45, 289-290.
Received 20 June 1991; revised version received 21 November 1991

Vous aimerez peut-être aussi