Vous êtes sur la page 1sur 17

451

Journal of Occupational and Organizational Psychology (2003), 76, 451467


2003 The British Psychological Society
www.bps.org.uk

Development and validation of the Job Security


Index and the Job Security Satisfaction scale:
A classical test theory and IRT approach
Tahira M. Probst*
Washington State University Vancouver, USA
This study describes the development and validation of two scales measuring job
security: the Job Security Index (JSI), measuring an individuals cognitive appraisal of
the future of his or her job with respect to the perceived level of stability and
continuance of that job, and the Job Security Satisfaction (JSS) scale, measuring
employee satisfaction with a perceived level of job security. Item response theory
and classical test theory analyses indicate that the scales are highly reliable and
exhibit good discriminant and criterion-related validity. Future researchers are
advised to apply con rmatory factor analysis (CFA) to the scales to con rm their
factor structures.

A continuing problem in the field of job security research has been the lack of a
comprehensive definition and measurement of job security. Many studies fail to
define the construct and/or use one- or two-item indicators to measure job security.
Commonly, participants are asked to indicate the likelihood of losing their job in the
next year or if they expect a change in their employment for the worse (e.g.
Buessing, 1986; Dooley, Rook, & Catalano, 1987; Roskies, Louis-Guerin, & Fournier,
1993). Other researchers partition their sample according to those in organizations
with low job security and those in organizations with high job security, this being
defined by the researcher (e.g. Buessing, 1986).
Greenhalgh and Rosenblatt (1984) were among the first researchers to provide a
rigorous definition of the job insecurity construct as well as explicate some important organizational outcomes of the phenomenon. They defined job insecurity as
perceived powerlessness to maintain desired continuity in a threatened job situation
(p. 438). Based on this definition, Ashford, Lee, and Bobko (1989) developed a 57-item
measure known as the Job Insecurity Scale (JIS), which measured job insecurity by
assessing: (1) the range of work situation features that could be in jeopardy, (2) the
valence of each such feature, (3) the subjective probability of losing each feature, and
(4) the number of sources of threat. Thus, job insecurity =[(S importance of job
featurelikelihood of losing job feature)+ (S importance of job losslikelihood of
job loss)] perceived powerlessness to resist threat.

*Requests for reprints should be addressed to Tahira M. Probst, Assistant Professor, Department of Psychology,
Washington State University Vancouver, 14204 NE Salmon Creek Avenue, Vancouver, WA 98686, USA (e-mail:
probst@vancouver.wsu.edu).

452

Tahira M. Probst

Although this measure has seen some use, there are several potential drawbacks to
defining and measuring job insecurity as such. First, it is unclear how conceptually
distinct perceptions of loss of job features are from loss of employment. For
example, Roskies and Louis-Guerin (1990) concluded that the antecedents and consequences of threat of demotion and threat of termination were sufficiently similar to be
grouped together. Second, when conceptualizing job security, it is important to clearly
distinguish dimensions of the construct itself from those variables that moderate the
relationships between the construct and outcome variables of interest. Although perceived powerlessness is undoubtedly an important variable in the study of job security,
it is more likely to be a moderator of job security, rather than an aspect of job security
per se. Similarly, the definition implies that an individual experiences job insecurity
only if the individual desires continuity in that work role. It would seem, however, that
an individual experiences job insecurity regardless of desire for continuity, but that
desire for continuity would moderate the impact that job insecurity has on outcome
variables. In support of this, past research has shown that job importance moderates
the relationship between perceived job security and several individual outcomes such
as psychological distress, physical health, and job satisfaction (Orpen, 1993; Probst,
2000). Therefore, by including job importance in the definition of job security, any
information regarding the interactive effects between job importance and perceptions
of job insecurity on the outcome variables of interest is lost.
Based on the preceding arguments, the purpose of this paper was to develop an
alternative theoretically grounded and empirically tested definition and measurement of job security. This study defines job security as the perceived stability and
continuance of ones job as one knows it (Probst, 2002). While proposing that job
security is essentially a global construct, this definition does allow that ones perception of job security can be influenced by both continuance of ones job and stability
with respect to desired features of ones job. In contrast to other definitions of job
security (e.g. Ashford et al., 1989; Greenhalgh & Rosenblatt, 1984), this definition does
not attach or include in its definition any attitudinal or affective reactions to perceived
job insecurity. Job insecurity is not predicted to occur only when an individual
perceives that the future of his or her job is unstable and negatively reacts to it. Rather,
it is proposed that job insecurity exists when the future of the individuals job is
perceived to be unstable or at risk. Thus, the construct is meant only to encapsulate
the perceived level of stability and continuance of ones job. Restricting the definition
and measurement of job security to perceptions of job security allows researchers to
then explore the moderators of the relationship between the individuals perception of
job insecurity and his or her evaluative and affective responses to that perception.
In this paper, two measures related to job security were developed. The first, the Job
Security Index (JSI), was designed to assess an individuals cognitive appraisal of the
future of his or her job with respect to the perceived level of stability and continuance
of that job. The second measure, the Job Security Satisfaction (JSS) scale, was designed
to assess satisfaction with job security. Whereas the JSI was designed to assess
perceptions of job security, the JSS scale was designed to capture an individuals
attitudes regarding that level of job security (i.e. their affective reactions).
Two distinct measures are necessary because ones perceptions of job security
although likely to be related to are not the same as ones affective reaction to that job
security. For example, one could envision two employees occupying identical
positions with identical levels of job security. However, one would not necessarily
expect that these two individuals would have identical affective reactions to these

Measures of job security

453

Figure 1. A conceptual model of the antecedents and consequences of job security. Adapted from
Probst (2002).

perceptions. Their affective reactions could be influenced by a number of variables


ranging from their financial dependence on the job to their personality characteristics.
Probst (2002) has developed a conceptual model of the antecedents and
consequences of job security (see Fig. 1). Although an in-depth discussion is beyond
the scope of this paper, one can see the distinct relationships that would be predicted
between job security perceptions and job security satisfaction and their unique
antecedents and consequences. For example, organizational change such as downsizing is expected to influence ones perceptions of job security. However, such
change would not be expected to directly influence ones satisfaction with job
security. Conversely, the effects of job security perceptions on outcomes such as
physical and mental health, job stress, and turnover intentions are expected to be fully
mediated by ones job security satisfaction. Thus, by using the conceptual model
developed by Probst (2002) to formulate hypotheses, one can assess the validity of the
proposed scales in a nomological network framework a
` la Cronbach and Meehl
(1955).

Method
Participants and procedure
A sample group of 500 state government employees from five different health and
human services agencies were randomly selected to participate in this study. These
state agencies were in the process of being merged into one umbrella human services
organization. As a result of this restructuring, possible job changes included moving
offices, changing supervisors, being demoted, learning new job-related technologies,
a reduction in job status, and/or new work tasks. Within the same agency, some
employees were in danger of losing key features of their job, or their current position,
while others remained essentially unaffected by the reorganization.
Of the 500 randomly selected employees, 283 agreed to participate in the study and
were given a survey to complete during working hours. The modal age of participants
was 4549 years; 64% of participants were women and 30% belonged to racial or
ethnic minorities; 34% of the respondents were managers and the modal level of

454

Tahira M. Probst

education was some college. The mean tenure of participants within the organization
was 13.06 years (SD= 8.8 years). These sample demographics were representative of
the overall organization.

Measures
The measures contained in the survey are described below.
Job Security Index (JSI)
The JSI consists of a series of adjectives or short phrases describing the future of
an employees job, e.g. my job is almost guaranteed, can depend on being here,
permanent position if I want it, insecure. Respondents indicated on a 3-point scale
(yes, ?, no) the extent to which the adjectives or phrases described the future of their
job. A 3-point response format was chosen because of numerous past research on
similar scales (e.g. the JDI) showing that this format is easy even for respondents with
very low reading ability to comprehend and discriminate among the categories
(Roznowski, 1989; Smith, Kendall, & Hulin, 1969). Responses were scored such that
higher numbers reflected greater perceived security. Positive and negative items were
intermixed in equal numbers on the page to minimize any response set.
The initial item pool was developed by selecting appropriate phrases from past
research on job security. The item pool was supplemented by adjectival phrases that
could also be used to describe the perceived security of ones job. This pool of phrases
was then further developed through a thesaurus search of related concepts. These
searches resulted in an initial pool of 40 phrases. After receiving feedback from various
colleagues and subject-matter experts, a number of phrases were removed from consideration. For example, temporary was not included in the final scale as this also
describes the legal status of a job (e.g. a temporary position versus a permanent
position). Another phrase, protected was dropped because it was thought this might
lead to different interpretations by different samples (e.g. government workers vs.
private sector employees or union vs. non-union positions). In choosing the final set of
items, great care was taken only to include items that were descriptive not evaluative
of the future of ones job. Following this process, 18 descriptive phrases describing
the future of ones job were selected for final inclusion in the scale (see Table 4 for a
complete list of items). The scale had a Cronbachs a of .97 and a mean of 1.51
(SD=1.05).
Job Security Satisfaction (JSS)
The JSS consists of a series of adjectives or short phrases describing the various
evaluative responses one might have to a perceived level of job security. Although the
evaluative response was targeted, adjectives or short phrases such as sufficient
amount of security, unacceptably low and makes me anxious were developed
to span the descriptiveevaluative continuum in an attempt to avoid any general
affectivity bias. Phrases that were purely descriptive, however, were not included,
as these types of items would be better suited for the JSI. Based on a search of the
available job security literature, discussions with subject-matter experts, and a thesaurus search, an initial pool of 35 phrases was developed. After receiving feedback from
various colleagues and subject-matter experts, a number of phrases were removed
from consideration. For example, daunting, aggravating, and exasperating were
removed owing to concerns that respondents with lower levels of reading ability might

Measures of job security

455

have difficulties with the words. Following this process, 20 descriptive phrases
describing ones job security satisfaction were selected for final inclusion in the scale
(see Table 5 for a complete list of items). The JSS had a Cronbachs a of .96 and a mean
of 1.71 (SD=.95).
Positively and negatively worded items were intermixed in equal numbers on the
page to minimize any response set. In an identical fashion as the JSI, respondents were
again instructed to indicate using a 3-point scale (yes, ?, no) the extent to which each
word or phrase described their job security. Responses were scored such that higher
numbers reflected greater satisfaction with ones job security.
Discriminant validation measures
In order to assess whether the JSS measures a unique component of job satisfaction
and, thus, to provide discriminant validity evidence for the JSS, respondents completed
the five, 9-item Job Descriptive Index scales (JDI; Smith et al., 1969), as revised by
Roznowski (1989). Using a 3-point scale, the JDI measures satisfaction with
co-workers (M =2.32, SD=0.77, a=.87), pay (M =1.61, SD=0.86, a=.84), promotions
(M =0.87, SD=0.88, a=.89), supervision (M =2.25, SD =0.83, a=.88), and the work
itself (M=1.83, SD= 0.98, a=.90).
Criterion-related validation measures
In order to provide evidence of criterion-related validity, several additional measures
were completed. The physical health of respondents was measured using the Health
Conditions Index (HCI), a yes/no checklist of health and physical symptoms adapted
from the Cornell Medical Checklist (Brodman, Erdman, Lorge, & Wolff, 1949). This
self-report scale asks respondents to indicate whether or not they have experienced 13
different health conditions (e.g. back problems, severe headaches, frequent colds) in
the past year. The HCI had a Cronbachs a of .75 and a mean of 2.39 (SD =2.42).
Psychological distress was measured using 15 items from the Mental Health Index
(Veit & Ware, 1983). The MHI measures emotional ties, anxiety, depression and loss of
behavioral/emotional control on a 6-point Likert-type scale. The MHI had a Cronbachs
a of .93 and a mean of 2.32 (SD =0.76).
Finally, 10 items from Matteson and Ivancevichs (1982) Stress Diagnostic Survey
(SDS) were used to measure the degree of job and organizational stress experienced by
the respondents. These items tap constructs such as role ambiguity, role conflict, work
overload and time pressure on a 5-point Likert scale. The SDS had a Cronbachs a of .76
and a mean of 2.63 (SD=0.66).
Other variables
Participants were asked to complete the 13-item Neutral Objects Satisfaction
Questionnaire (Weitz, 1952), which measures respondents negative affectivity on a
3-point scale, where higher numbers indicate a greater tendency for negative
affectivity (M= 1.56, SD= 0.32, a=.77). Employees were also asked to indicate in a
checklist format the number of job changes (out of 11 listed) that they had
experienced as a result of the reorganization (M =3.71, SD=2.18, a=.70). Finally,
respondents were asked how many years they had been with the organization.
Data analysis
Con rmatory factor analysis
In order to provide construct validity evidence for the hypothesized underlying
structure of the JSI and JSS scales, confirmatory factor analyses (CFA) were

456

Tahira M. Probst

performed on the set of JSI and JSS items using LISREL 8 (Jo
reskog & Sorbom, 1993).
The structural equations for the latent JSI and JSS constructs in the confirmatory factor
analysis were specified as follows: x=L x x+ d, where x=the underlying latent construct,
x= the observed indicators of x, L x = coefficients relating x to x, and d=measurement
errors for x.
In order to assess the fit of the hypothesized underlying factor structure, several fit
indices were examined. The first is the c2 statistic. However, c2 is highly sensitive to
sample size, thus frequently resulting in a rejection of the model when the model may
not in fact be off-target (Mulaik et al., 1989). Therefore, a general rule of thumb is to
divide the c2 statistic by the number of degrees of freedom (df). If the resulting value is
greater than 3.0, the model may be rejected due to poor fit (Wheaton, Muthen, Alwin,
& Summers, 1977). In addition, when comparing nested models, the sequential c2
statistic is useful in determining whether the fit of the model significantly improves
after accounting for the difference in model degrees of freedom (Kelloway, 1998,
pp. 3637).
Two additional useful indices are the goodness-of-fit index (GFI) and the adjusted
goodness-of-fit index (AGFI). Because the GFI can be improved by freeing up more
parameters in a model, the AGFI is often cited along with the GFI, since it is designed
to compensate for this increase in fit due to an increase in free parameters. The
non-normed fit index (NNFI) is also useful when assessing the goodness of fit of a
model. A general rule of thumb states that models with NNFIs below .90 can be
improved substantially (Bentler & Bonett, 1980).
Analysis of the residuals representing the difference between the elements of the
unrestricted sample covariance matrix and those of the hypothetical model covariance
matrix for the observed variables is also informative. The root mean square residual
(RMSR) is a measure of the average of fitted residuals. The RMSR is useful in that it can
be used to compare the fit of two different models for the same data (Jo
reskog &
Sorbom, 1993). An additional index of lack of fit is the root mean square error of
approximation (RMSEA). In general, RMSRs and RMSEAs around .05 indicate a good fit
of the model to the data.
Exploratory factor analysis
Following the CFA, two additional exploratory factor analyses were also performed in
the development and validation of the JSI and JSS scales. First, in order to provide
an assessment of the underlying structure of the JSI and JSS items, an exploratory
factor analysis with oblique rotation (given the expectation that the scales would be
correlated) was performed on the entire set of 38 items. Second, to provide evidence
of discriminant validity for the JSS scale, the 20 JSS items were entered into an
exploratory factor analysis with a varimax (orthogonal) rotation with the items from
the five JDI scales.
Item Response Theory (IRT) analyses
In addition to assessing the psychometric properties of the scales items by using
classical theory statistics (e.g. item-endorsement indices, point biserial correlations and
factor loadings), item response theory statistics were also computed and assessed using
BILOG (Mislevy & Bock, 1989). Because there is no guessing on the JSI or JSS scales,
the scale items were modelled using a two-parameter IRT model. For each item,
an item discrimination (slope or ai) and an item difficulty (threshold or bi) were
estimated. The item discrimination index can range from 0 to >3; however, ai

Measures of job security

457

parameters >3 are perhaps too discriminating. The item difficulty index indicates on
a normalized scale ranging from 3 to +3 the perception (for the JSI) or satisfaction
(for the JSS) level needed for 50% of respondents to endorse the item. A good scale
would have items that span that range.
Other analyses
Multiple regression analyses were performed to assess the evidence for discriminant
and criterion-related validity by assessing various relationships predicted by the conceptual model of the antecedents and consequences of job security discussed earlier
(Probst, 2002). Additional construct validity evidence was assessed by looking at the
correlations between negative affectivity and the two job security constructs.

Results
Table 1 contains the intercorrelations of the job security scales and the validation
measures. As can be seen, the JSI and JSS were strongly correlated at r=.71. Job
security perceptions were also modestly correlated with the JDI measures (excepting
the supervisor satisfaction scale), physical and mental health, job stress, job changes,
and organizational tenure. Notably, respondents perceptions of job security were
unrelated to their levels of negative affectivity, indicating that their job security
perceptions were not a function of any predisposition towards negative affectivity.
As expected, the JSS scale was more strongly correlated with the JDI measures than
was the JSI. All correlations were significant and positive, ranging from .18 to .37.
In addition, greater job security satisfaction was significantly negatively related to
physical and mental health conditions and job stress. It is worth noting that job
security satisfaction was not significantly correlated with the number of job changes
experienced by the respondents or the length of time they had been with the organization. Finally, there was a negative relationship between the JSS scale and negative
affectivity r= .14, p<.05), indicating that a predisposition towards negative affectivity
is related to greater job security dissatisfaction. Thus, negative affectivity is somewhat related to an individuals satisfaction with his or her job security, but not the
perception of how secure that job is.
Evidence of JSI and JSS discriminant validity
The first discriminant validity analyses were conducted to demonstrate that, although
related, job security perceptions and job security satisfaction are unique constructs. In
order to demonstrate this, a confirmatory factor analysis followed by an exploratory
factor analysis was performed on the item-level data.
Con rmatory factor analysis
An initial confirmatory factor analysis of the JSI and JSS scale items using LISREL 8
(Joreskog & Sorbom, 1993) was conducted in order to assess the support for two
distinct constructs. Two-factor analytic models were fitted to the data. The first model,
a one-factor model, posited that there is a general factor of security that accounts for
variation on both scales. The second model, a two-factor model, posited that (a) there
are two factors, perceptions of security and satisfaction with security, and (b) both are
necessary to fit the data well. Results for the confirmatory factor analyses are shown in
Table 2.

Job security perceptions


Job security satisfaction
Co-worker satisfaction
Pay satisfaction
Promotion satisfaction
Supervisor satisfaction
Work satisfaction
Physical health conditions
Psychological distress
Job stress
Job changes
Organizational tenure
Negative affectivity

.17**
2 .04

.71**
.16**
.18**
.27**
.06
.12*
2 .20**
2 .19**
2 .20**

.18**
.29**
.37**
.21**
.25**
.34**
.37**
.28**
.14*
.03
.14*

.23**
.22**
.26**
.37**
.14*
.22**
.28**
.03
.10
.24**

.24**
.12*
.27**
.18**
.10
.13*
.04
.14*
.20**

.28**
.34**
2 .21**
2 .24**
2 .35**
.02
.01
.02

.19**
2 .08
2 .17**

2 .47**

2 .26**

.24**

2 .27**

.11
2 .18**

2 .00

2 .26**

2 .29**

2 .20**

Note: *p<.05, **p<.01; The JSS scale used is the nal 18-item version developed subsequent to the CFA and EFA analyses.

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.

Variables

Table 1. Intercorrelations of job security scales and validation measures

.53**
.35**
.12*
2 .05
.08

.34**
.10
2 .02
.23**

.03
.08
.17**

10

.02

.12*

2 .13*

11

.03

12

458
Tahira M. Probst

Measures of job security

459

Table 2. Goodness-of-t indices for job security scales con rmatory factor analyses
Model
One-factor model
Two-factor model

2,295.13
1,513.40

df

v 2 /df

GFI

AGFI

NNFI

RMSEA

RMSR

665
664

3.45
2.28

.52
.72

.46
.69

.77
.88

.10
.07

.08
.06

Note: GFI= goodness-of- t index; AGFI=adjusted goodness-of- t index; NNFI=non-normed t index;


RMSEA=root mean square error of approximation; RMSR=root mean square residual; One-factor model=Job
Security Index and Job Security Satisfaction scales combined; Two-factor model=Job Security Index and Job
Security Satisfaction Scales separate.

The goodness-of-fit indices strongly suggest that the two-factor model better
accounts for the data than the one-factor model. In each instance the fit indices are
significantly improved in the two-factor model. The GFI increases from .52 to .72, the
AGFI from .46 to .69, and the NNFI from .77 to .88. In addition, the RMSEA decreases
from .10 for the one-factor model to .07 in the two-factor model, while the RMSR is
reduced from .08 to .06.
Although the fit indices suggest the two-factor solution provides a significantly
better fit to the data, the resulting fit indices are nonetheless lower than would be
desired according to conventional standards. Therefore, in order to further develop the
scale, an exploratory factor analysis was performed.
Exploratory factor analysis
Using a principal components extraction with an oblique rotation (to account for the
correlation between the two scales), an exploratory factor analysis was performed on
the data using SPSS to assess whether two factors would be extracted from the data. In
fact, four factors were extracted, accounting for a total of 68.68% of the variance. The
first factor was a clear factor for the JSI items, with the majority of the JSI items loading
in the .60.90 range. The lowest factor loading was .38 (permanent position if I want
it). The second and third factors consisted entirely of JSS items, with the negatively
phrased JSS items loading onto the second factor and the positively worded JSS items
loading onto the third factor. No JSI items cross-loaded onto either JSS factor above .20.
A minor fourth factor (accounting for little over 2% of the variance with an eigenvalue
of 1.03) was generally uninterpretable, consisting primarily of two JSS items (never
been more secure and excellent amount of security) that also did not cross-load onto
either of the JSS factors nor the JSI factor. On the basis of this analysis, these two
problematic JSS items were removed from the scale.
Additional discriminant and criterion-related validity analyses
Additional discriminant and criterion-related validity evidence can be found in the
regression analyses presented in Table 3.
Consequences of job security satisfaction
According to the conceptual model developed by Probst (2002), only if one is
dissatisfied with ones level of job security should one experience stress as a result. In
fact, stress levels were significantly predicted by job security satisfaction (b = .28,
p<.001), but were not predicted by job security perceptions (b = .01, n.s.). Similarly,

460

Tahira M. Probst

Table 3. Criterion-related and discriminant validity regression analyses


Variable
Job stress
Job security perceptions
Job security satisfaction
F(2, 269)=11.77, p<.001; R2adj =.07

p
b

2 .01

ns
.001

.07

ns
.001

.12

ns
.001

2 .28

Physical health conditions


Job security perceptions
Job security satisfaction
F(2, 271)=17.95, p<.001; R2adj =.11

2 .39

Psychological distress
Job security perceptions
Job security satisfaction
F(2, 272)=23.13, p<.001; R2adj =.14

2 .46

according to the conceptual model, one might expect an individual to experience a


higher incidence of physical and mental health conditions as an outcome of job
insecurity only if one is dissatisfied with that insecurity. As expected, physical health
conditions were predicted by satisfaction with job security (b = .39, p<.001), but
were not predicted by perceptions of security (b =.07, n.s.). In a similar fashion,
mental health was predicted by satisfaction with job security (b = .46, p<.001), but
was not predicted by perceptions of security (b =.12, n.s.). These analyses suggest
again that the JSI and JSS are unique constructs. In addition, they provide criterionrelated validity evidence for the JSS.
Because the EFA results indicated that there are positive and negative factors for the
JSS, it is important to determine whether the predicted relationships between the JSS
and the relevant criteria are comparable for both JSS factors. Therefore, zero-order
correlations between each subscale of the JSS and the three criteria were next examined. These results indicate that both subscales of the JSS have highly similar relationships with job stress (JSS-negative r= .30; JSS-positive r= .24), physical health
(JSS-negative r= .35; JSS-positive r= .29), and mental health (JSS-negative r= .35;
JSS-positive r= .38). Thus, regardless of the positive or negative wording of the JSS
items, these factors both display the predicted relationships with the criteria of
interest. Because the positive and negative factors correlate equally with these criteria,
it appears that the positive and negative factors can perhaps best be viewed as
nuisance factors rather than difficulty factors (e.g. Nunnally & Bernstein, 1994) and can
be meaningfully combined into a single scale.
Predictors of job security perceptions
As predicted by the conceptual model (Probst, 2002), variables that might be expected
to influence perceptions of job security, but not be directly related to satisfaction
with job security, operated in the anticipated manner. For example, individuals who
reported having more job changes as a result of the current organizational restructuring also reported perceiving lower job security r= .14, p<.05). However, the degree
to which the individuals were affected by the restructuring was unrelated to their

Measures of job security

461

satisfaction with their job security (r= .04, n.s.). In addition, individuals who had
been with the organization for a greater period of time perceived their jobs to be more
secure (r= .17, p<.01), but were not necessarily more satisfied with that security
(r= .01, n.s.).1 Thus, two variables that were expected to predict job security
perceptions, but not job security satisfaction, operated in the expected manner.
Discriminant validity evidence for the JSS as a unique facet of job satisfaction
It is also important to examine whether job security satisfaction has the potential to
significantly contribute to our understanding of job attitudes or if attitudes regarding
job security are adequately captured by existing measures of job satisfaction.
Job satisfaction has traditionally been conceptualized as consisting of five facets:
(a) satisfaction with the work itself, (b) co-worker satisfaction, (c) pay satisfaction,
(d) promotion satisfaction, and (e) supervisor satisfaction (Smith et al., 1969). Job
security satisfaction had been shown by Harrison (1961) to be a facet of job satisfaction
using factor analytic techniques and was originally conceptualized as a facet of job
satisfaction by the creators of the Job Descriptive Index (C. L. Hulin, personal
communication, 14 February 1996). However, there was little evidence of discriminant
validity for the construct. Looking back, this was probably because there would have
been little or no variance on the variable at the time the JDI was developed. The era
of massive layoffs had not yet begun and lifetime security was taken more-or-less
for granted.
The correlation matrix in Table 1 shows that while job security satisfaction is
significantly related to the majority of the traditional job satisfaction facets, the correlations are low enough to provide support for the argument that the JSS is measuring a
unique construct. In addition, an exploratory principal components factor analysis
with varimax rotation of the items from the six satisfaction scales reveals six factors,
with the JSS items loading onto a single factor accounting for 17.2% of the variance
(rather than loading separately on positive and negative factors). The highest
cross-loading of a JSS item was .22 onto the work satisfaction factor.
Given that adequate construct, discriminant and criterion-related validity evidence
was demonstrated, the psychometric properties of the JSI and JSS were examined in
greater detail using both classical test theory and item response theory methodologies.
Psychometric properties of the JSI and JSS scale items
In the first analysis, the psychometric properties of the individual items and scales
were evaluated in terms of classical test theory. Classical test theory statistics for the JSI
are presented in Table 4. Scores for the JSI could range from 0 to 3. The mean was
1.51, SD= 1.05. The a reliability coefficient for the JSI was quite high, a=.97. The
proportion of items that were endorsed to indicate high security (an indicator of item
difficulty) ranged from .28 to .51. The itemtotal biserial correlation (an indicator
of item discrimination) ranged from 0.72 to 1.12. (Because itemtotal biserials are
estimates, they can assume values greater than 1. If this occurs frequently among items
in a scale, it may indicate that the scale items are more homogeneous than would be
desirable.)
1

Notably, the non-signicant relationship between these predictors and job security satisfaction were also demonstrated for
both the positive and negative JSS subscales again suggesting these factors are nuisance factors rather than having
substantive meaning.

462

Tahira M. Probst

Table 4. Classical test theory statistics and item response theory parameters for the JSI

Item
1. Sure
2. Unpredictable
3. Up in the air
4. Secure
5. Stable
6. Questionable
7. Unknown
8. Well established
9. My job is almost guaranteed
10. Uncertain
11. Can depend on being here
12. Future is vague
13. Unclear
14. Permanent position if I want it
15. Certain
16. This job might not be around too long
17. Unspeci ed
18. Insecure

Classical test
theory statistics

Item response
theory parameters

rb

ai

bi

.438
.368
.509
.428
.466
.437
.395
.439
.277
.386
.369
.396
.372
.388
.356
.425
.410
.459

1.04
1.00
0.91
1.02
0.94
1.09
1.02
0.88
0.84
1.12
1.00
1.03
1.07
0.72
1.01
0.95
0.98
1.08

1.72
1.49
1.37
1.58
1.36
2.02
1.57
1.19
1.08
3.02
1.69
1.93
2.32
0.84
1.72
1.45
1.65
3.02

0.168
0.456
2 0.094
0.207
0.083
0.158
0.343
0.191
0.902
0.332
0.432
0.310
0.401
0.470
0.496
0.238
0.280
0.070

Note: p=proportion correct; rb =itemtotal biserial correlation; ai =item discrimination; bi =item difficulty.

The classical test theory results for the JSS scale are presented in Table 5. Again,
scores for the JSS could range from 0 to 3. The mean was 1.71 and the SD was 0.95.
The a coefficient for the JSS was also quite high, a=.96. The proportion of items
positively endorsed to indicate job security satisfaction ranged from .19 to .82. The
itemtotal biserials ranged in value from 0.42 to 1.03.
Based on classical test theory indices, one can conclude that the JSI and JSS are
highly reliable scales. There is a fair amount of variation in participant endorsement of
the JSI items, indicating the items are of varying popularity. The variation in participant
endorsement of the JSS is even greater, again indicating that the scale performs well at
measuring varying degrees of job security satisfaction. Itemtotal biserial correlations
are generally quite high, indicating high levels of discrimination.
Because classical test theory indices are sample-dependent, it is also useful to
estimate and evaluate item parameters based on item response theory methods (Hulin,
Drasgow, & Parsons, 1984). Therefore, the items from each scale were modelled using
Mislevy and Bocks BILOG program (1989) to examine estimated item difficulties and
item discrimination levels. Item response theory results for the JSI are presented in
Table 4. As can be seen, the majority of the JSI items are quite discriminating
(ai parameters range from 0.84 to 3.02). It should be noted, however, that items with
ai parameters >2.50 may be too discriminating, meaning that they provide very precise
measurement but only for a rather narrow range of q levels. Two items, item 10
uncertain and item 18 insecure, have ai parameters >3.
The bi parameters in item response theory provide an indication of item difficulty. It
corresponds to the q level at which the probability of positively endorsing the item is

463

Measures of job security


Table 5. Classical test theory statistics and item response theory parameters for the JSS scale
Classical test
theory statistics
Item
1. Never been more secure
2. Makes me tense
3. Satisfactory
4. Nerve-wracking
5. Sufficient amount of security
6. Cause for concern
7. Acceptable
8. Discouraging
9. Inadequate
10. More secure than most in my job or profession
11. Worrisome
12. Looks optimistic
13. Makes me anxious
14. Upsetting how little job security I have
15. Excellent amount of security
16. All right
17. Stressful
18. Positive
19. Unacceptably low
20. Troubling

Item response
theory parameters

rb

ai

.191
.544
.537
.615
.524
.461
.569
.544
.526
.294
.567
.377
.561
.658
.204
.602
.481
.412
.637
.823

0.69
0.85
0.91
0.88
0.85
0.93
1.02
0.94
1.00
0.42
0.96
0.68
0.84
0.91
0.77
0.87
0.84
0.92
0.86
1.03

1.18
1.36
1.65
1.57
1.41
1.84
2.53
1.90
2.56
0.54
2.02
0.93
1.38
1.70
1.59
1.41
1.37
1.83
1.45
2.93

bi
1.202

2 0.136
2 0.101
2 0.358
2 0.070
0.134

2 0.172
2 0.118
2 0.050
1.132

2 0.184
0.480

2 0.188
2 0.501
1.033

2 0.331
0.075
0.289
2 0.446
2 0.118

Note: p=proportion correct; rb =itemtotal biserial correlation; ai =item discrimination; bi =item difficulty.

.50. The bi parameters range from 0.094 to 0.902, suggesting there is a fair range of
item difficulties present in the scale. However, these parameters also indicate that
it might be useful in the future to develop somewhat more difficult items which
discriminate among individuals who have lower than average perceptions of job
security.
The item response theory results for the JSS scale are presented in Table 5. Overall,
the items appear to discriminate quite well among individuals at varying levels of q.
The ai parameters range from 0.54 to 2.93. Of the 20 items, only item 10 (more secure
than most in my job or profession) had a somewhat lower than desired discrimination
parameter (ai =.54), meaning that it does not discriminate very well among individuals
with different levels of job security satisfaction. The bi or item difficulty parameters
range from 0.501 to 1.202, meaning that the items are likely to be endorsed with
substantially different probabilities by individuals at varying levels of the latent trait.
For example, item 14 upsetting how little security I have is an easy item, such that
individuals at q= 0.501 have a probability of .5 of endorsing the item. In other words,
unless an individual is half a standard deviation below average on the latent trait, that
individual has a fairly low probability of endorsing that item. On the other hand, item
1, never been more secure, is a rather difficult item, meaning it is likely to be
positively endorsed only by individuals with relatively high job security satisfaction
(e.g. individuals at q=1.202 have a probability of .5 of positively endorsing this item).

464

Tahira M. Probst

Discussion
The purpose of this study was to develop two measures of job security: (1) the
JSI, a measure of perceived job security, and (2) the JSS scale, a measure of satisfaction
with ones job security. Based on the analyses reported, it is argued that several
conclusions can be drawn about the psychometric properties of the scales. First,
perceptions of job security appear to be unique from ones evaluative response to
ones job security. The data suggest that the JSI and the JSS scale, while moderately
correlated, nonetheless measure unique constructs each with its own unique antecedents and consequences. Second, job security satisfaction appears to be a sixth
component of job satisfaction, accounting for unique variance, and should be included
in future studies where inclusion would contribute to a greater understanding of the
topic being researched. Third, the JSI and JSS exhibit good psychometric properties
based on classical test theory statistics (e.g. high internal consistency, high itemtotal
biserial correlations, and a fairly wide range of item difficulties). Finally, item response
theory analyses reveal that the JSI and JSS items are quite discriminating and measure
well at varying degrees of item difficulty.
Suggestions for scale revision
Although the JSI and JSS exhibit satisfactory psychometric properties, there is room for
improvement. First, based on the exploratory factor analysis, two JSS items needed to
be deleted from the scale never been more secure and excellent amount of
security. These items formed a fourth minor factor and did not load well onto either
of the JSS factors. One possible explanation for this phenomenon can be found by
examining the results in Table 5. Both of these items are difficult items, meaning that
they were rarely endorsed, and one had to have a high level of job security satisfaction
before one would agree with the items. Given the organizational changes that were
occurring in the organization where the data were collected, these results may not be
surprising. Thus, additional data should be collected on these scales in organizations
where fewer organizational changes are pending. This would also increase the external
validity of the scale by expanding it to populations of employees other than those in
the public sector.
Based on the classical test theory and IRT results, there may also be a few items on
both scales that discriminate too well. For example, on the JSS scale, troubling is an
item that may only provide good discrimination for a very narrow range of job security
satisfaction. On the JSI, uncertain and insecure are also perhaps too discriminating.
The item response theory results also suggest that an attempt should be made to
develop items which discriminate well among individuals with lower than average
perceptions of security.
One other troubling finding was the separate factors exhibited for the negatively vs.
positively worded JSS items. These results potentially question the construct validity
of the JSS. However, subsequent analyses suggest that these JSS subfactors are more
likely owing to nuisance factors rather than substantive differences in the two sets
of items. Both negatively and positively worded items exhibited similar relationships
with mental health, physical health conditions, and job stress. In addition, both
subscales exhibited similar relationships with other study variables. Nonetheless,
future researchers should apply a confirmatory factor analysis to the revised scale in
order to confirm its factor structure.
Finally, additional analyses should be conducted to develop appropriate and psychometrically sound shortened versions of these scales. In organizational research, survey

Measures of job security

465

space is often at a premium. Given the extremely high internal consistency of these
scales and psychometric properties of the items, shorter versions should be able to be
developed while retaining the scales validity.

Implications for future research


Organizational restructuring in the form of corporate down-sizing, mergers and
acquisitions, plant closures, and workforce reorganizations affect millions of workers
each year. Annually, approximately 500,000 US employees can expect to lose their
jobs as a result of these transitions (Simons, 1998). In 2001 alone, over 1.4 million jobs
were lost from the US economy. In Europe, over 5 million jobs were lost in the early
1990s, with job losses exceeding job creation (European Commission, 1997) and
European unemployment has stubbornly remained above 8% for a number of years
(Eurostat, 2001).
Researchers have suggested that these organizational trends herald a new and
changing psychological contract (Rousseau, 1995) between employer and employee.
Empirical research suggests that this new contract may go against individual needs for
connectedness and affiliation (Kets de Vries & Balazs, 1997). Thus, researchers (e.g.
Hui & Tan, 1996) have called for more empirical research to provide a better understanding of the ways in which employees are affected by todays rapidly changing and
less stable work environment.
In order to study the phenomenon of todays job insecurity and the changing
psychological contract, one must have well-developed and validated measures of the
relevant constructs. In particular, one must have very clearly defined constructs and
measures that appropriately operationalize these constructs. This study was an attempt
to develop two such measures. The first, a measure of job security perceptions, is
intended for use to determine an individuals level of perceived job security. These job
security perceptions are expected to be influenced by organizational- (e.g. announced
layoffs) and individual-level (e.g. organizational tenure) variables. The second measure
assesses ones affective reactions to a perceived level of job security (i.e. job security
satisfaction). Research has shown (e.g. Probst, 2002; Probst & Brubaker, 2001) that this
variable is predictive of a host of individually (e.g. physical and mental health outcomes) and organizationally relevant (e.g. turnover and safety) outcomes. It is hoped
that these theoretically grounded and empirically validated measures of job security
will help advance the study of employees in todays rapidly changing workforce.

References
Ashford, S., Lee, C., & Bobko, P. (1989). Content, causes, and consequences of job insecurity: A
theory-based measure and substantive test. Academy of Management Journal, 32, 803829.
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of
covariance structures. Psychological Bulletin, 88, 588606.
Brodman, K. A., Erdman, J., Lorge, I., & Wolff, H. G. (1949). The Cornell Medical Index. Journal
of the American Medical Association, 140, 530534.
Buessing, A. (1986). Worker responses to job insecurity: A quasi-experimental field investigation.
In G. Debus, & H. W. Schroiff (Eds.), The psychology of work and organization
(pp. 137144). North Holland: Elsevier.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological
Bulletin, 52, 281302.

466

Tahira M. Probst

Dooley, D., Rook, K., & Catalano, R. (1987). Job and non-job stressors and their moderators.
Journal of Occupational Psychology, 60, 115132.
European Commission (1997). Commission draft for the joint employment report (Rev. 8).
Brussels: European Commission.
Eurostat (2001). Euro-zone unemployment unchanged at 8.7%. News release No. 16/2001.
Luxembourg: Eurostat.
Greenhalgh, L., & Rosenblatt, Z. (1984). Job insecurity: Towards conceptual clarity. Academy of
Management Review, 9, 438448.
Harrison, R. (1961). Cumulative communality cluster analysis of workers job attitudes. Journal
of Applied Psychology, 45, 123125.
Hui, C. H., & Tan, C. K. (1996). Employee motivation and attitudes in the Chinese workforce. In
M. H. Bond (Ed.), The handbook of Chinese psychology (pp. 364378). Hong Kong: Oxford
University Press.
Hulin, C. L., Drasgow, F., & Parsons, C. K. (1983). Item response theory: Applications to
psychological measurement. Homewood, IL: Irwin.
Joreskog, K. G., & So
rbom, D. (1993). LISREL 8: Users reference guide. Mooresville,
IN: Scientific Software.
Kelloway, E. K. (1998). Using LISREL for structural equation modeling. Thousand Oaks,
CA: Sage.
Kets de Vries, M. R., & Balazs, K. (1997). The downside of downsizing. Human Relations, 50,
1150.
Matteson, M. T., & Ivancevich, J. M. (1982). The how, what and why of stress management
training. Personnel Journal, 61(10), 768774.
Mislevy, R. J., & Bock, R. D. (1989). PC-BILOG 3. Mooresville, IN: Scientific Software.
Mulaik, S. A., James, L. R., Van Alstine, J., Bennett, N., Lind, S., & Stilwell, C. D. (1989).
Evaluation of goodness-of-fit indices for structural equations models. Psychological Bulletin,
105(3), 430445.
Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory. New York: McGraw-Hill.
Orpen, C. (1993). Correlations between job insecurity and psychological well-being among
White and Black employees in South Africa. Perceptual and Motor Skills, 76, 885886.
Probst, T. M. (2000). Wedded to the job: Moderating effects of job involvement on the
consequences of job insecurity. Journal of Occupational Health Psychology, 5(1), 6373.
Probst, T. M. (2002). The impact of job insecurity on employee work attitudes, job adaptation,
and organizational withdrawal behaviors. In J. M. Brett & F. Drasgow (Eds.) The psychology
of work: Theoretically based empirical research. Mahwah, NJ: Lawrence Erlbaum
Associates.
Probst, T. M., & Brubaker, T. L. (2001). The effects of job insecurity on employee safety
outcomes: Cross-sectional and longitudinal explorations. Journal of Occupational Health
Psychology, 6, 139159
Roskies, E., & Louis-Guerin, C. (1990). Job insecurity in managers: Antecedents and
consequences. Journal of Organizational Behavior, 11, 345359.
Roskies, E., Louis-Guerin, C., & Fournier, C. (1993). Coping with job insecurity: How does
personality make a difference? Journal of Organizational Behavior, 14, 617630.
Rousseau, D. M. (1995). Psychological contracts in organizations: Understanding written and
unwritten agreements. Thousand Oaks, CA: Sage Publications.
Roznowski, M. (1989). An examination of the measurement properties of the Job Descriptive
Index with experimental items. Journal of Applied Psychology, 74, 805814.
Simons, J. (1998, November 18). Despite low unemployment, layoffs soar corporate mergers
and overseas turmoil are cited as causes. Wall Street Journal, A2.
Smith, P. C., Kendall, L., & Hulin, C. L. (1969). The measurement of satisfaction in work and
retirement. Chicago: Rand McNally.
Veit, C. T., & Ware, J. E. (1983). The structure of psychological distress and well-being in general
populations. Journal of Consulting and Clinical Psychology, 51, 730742.

Measures of job security

467

Weitz, J. (1952). A neglected concept in the study of job satisfaction. Personnel Psychology, 5,
201205.
Wheaton, B., Muthen, B., Alwin, D. F., & Summers, G. F. (1977). Assessing reliability and stability
in panel models. In D. R. Heise (Ed.), Sociological methodology 1977 (pp. 84136).
San Francisco: Jossey-Bass.
Received 11 June 2001; revised version received 18 November 2002

Vous aimerez peut-être aussi