See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/267579464

BOTHMA, F.C. & ROODT, G. (2012). The

validation of the turnover intention scale.
Proceedings of the Pan Pacific Conference XXIX,
24 27 May, Haikou, China. pp. 231-235. (ISBN 193...
CONFERENCE PAPER JANUARY 2012

READS

221

2 AUTHORS:
Bothma Fc

Gerhard (Gert) Roodt

University of Johannesburg

University of Johannesburg

7 PUBLICATIONS 2 CITATIONS

213 PUBLICATIONS 280 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Gerhard (Gert) Roodt

Retrieved on: 10 October 2015

The validation of the Turnover Intention Scale

FC Bothma BothmaFC@telkom.co.za
G Roodt* groodt@uj.ac.za *corresponding author
PO Box 524, Auckland Park, 2006,
Department of Industrial Psychology and People Management
University of Johannesburg, South Africa
ABSTRACT
Much has been published about the retention of staff in the
recent past. While turnover intentions as a construct is
attracting increased research attention, there are seemingly
not many valid and reliable scales around to measure
turnover intentions. This study therefore focuses on the
validation of a shortened, 6-item version of the Turnover
Intention Scale (TIS-6). Several facets of validation will be
investigated, namely the reliability, as well as the construct,
predictive and discriminant validity of the scale. The study
was conducted amongst 2429 employees of an ICT sector
company as part of the first authors doctoral thesis. The
results of the study indicate that a shortened version of the
TIS could measure turnover intentions reliably ( = .80)
and validly. Leavers and stayers scored significantly
different on the TIS-6 and it also established significant
differences in respect of 8 of the 10 other theoretical
variables used in the study.
INTRODUCTION
The retention of staff is considered to be a pressing people
issue and consequently much has been published about it in
the recent past [23; 26; 34; 39; 53]. Turnover intention
(intentions to stay or leave the organisation) is an important
criterion variable in similar type of studies. The challenge
and importance of this study therefore is to investigate if
the TIS-6 can serve as a valid and reliable criterion variable
in future studies.
Although turnover intention is covered well in the
literature, the need remains to validate turnover cognition
scales [55]. The motivation for validating the shortened
version of the TIS-6 is that most other scales only use a
limited number of scale items. [43] observed that various
researchers have only used single item scales [28; 36] with
obvious metric limitations. According to [43] only a
limited number of other studies have used more than three
items in their instruments [11; 25; 41]. The main research
question of this study is: Is the TIS-6 a reliable and valid
scale? The objectives of the study are to investigate: (1)
the reliability; (2) the construct (factorial) validity; (3) the
predictive validity; and (4) the discriminant validity of the
TIS-6.
Turnover intention is seldom precisely defined in reported
studies. This is probably because people perceive the term
to be self-explanatory. Many researchers [29; 47; 48; 62]
argued that turnover intention is the final decision step
before a person actually leaves a workplace. Turnover
intention can be described as an individuals behavioural
intention (conation) to leave the employ of the organisation
[35]. In this study the definition of [64] is used, who define
turnover intention as: the conscious and deliberate
willfulness to leave the organisation (p. 262). The TIS
was developed as a conation (intention) of psychological
activities as conceptualised by [24].

Behavioural intention is a reliable determinant of actual

behavior [31; 49]. This implies that turnover intention can
be used as a valid proxy for actual labour turnover [31; 49;
64]. Turnover is the result of a coping strategy used by
employees to escape the current situation. It can be
permanent, when employees leave the employment
institution, or it can be characterised by horizontal mobility
when employees seek and accept transfers to other
departments [32]. [64] referred to turnover as the last in
a sequence of withdrawal cognitions (p. 262).
The decision to leave is influenced by many factors.
Perceived chances and ease of finding another job,
especially in tough economic conditions, the role of
mobility cognitions as well as individual differences in
search behaviour all contribute to an individuals turnover
intention [2; 3; 4; 12; 13; 15; 19; 20; 31; 38; 44; 47; 56;
67].
The Job Resources-Demands (JDR) Model [7; 8] provided
a plausible explanation why individuals may choose to
leave an organisation. In most studies that used the JDR
model, the path to turnover intentions is from job demands
that cause burnout. An indirect relationship between job
demands and turnover intentions is therefore proposed.
This idea is based upon studies which found that job
demands stimulate exhaustion (the opposite of
engagement), which causes turnover intentions [7; 8; 22;
58; 60]. These studies also indicated that the absence of
job resources was related to disengagement, which
increased turnover intentions. The finding that the absence
of job resources stimulated turnover intentions was also
supported in a study that did not use the JDR model [2].
The turnover phenomenon has significant cost
consequences for any organisation. Losing highly skilled
technical employees disrupts organisational functioning,
service delivery and administration. It also contributes to
hiring and training costs [53; 63]. These mentioned
consequences provide a sound rationale for this study. The
research design of the study is presented next.
RESEARCH DESIGN
Research approach
An empirical, quantitative research approach was followed
where a cross-sectional field survey was conducted that
generated primary research data.
Correlational data
analysis techniques were applied in the study, providing
plausible explanations for relationships in an ex post facto
way.
Research method
Participants and sampling
A census-based approach (where all members in the
sampling frame had an equal chance to participate) was
followed to survey the heterogenic population, up to middle
management (N = 23 134) in the employment of a large
South African Information and Communication

Technologies (ICT) sector company over a period of one

month. The response rate of 11% yielded a sample of 2 429
usable questionnaires.
Most participants were male (63.2%). The majority of
participants (44.1%) are white followed by black (26.3%),
coloured (16.3%) and Asian/Indian groups (13.3%). The
mean age of the participants was 40.07 years. Many were
from operational levels (54.9%) and were stationed in the
corporate region (24.9%) of the company.
Of the
participants, 40.7% had a Matric or lower qualification,
followed by 26.9% that possessed a National or National
Higher Diploma.
Measuring instruments
Turnover Intention Scale. Turnover intention (intention to
leave or stay) was measured with a 15-item scale initially
developed by [52]. Following literature recommendations
[49], behavior intention should be measured within a
reasonable time-frame, and for this study a six-month
period [49] was used. A total of six items were adapted
from [52]s TIS scale. Owing to space limitations no
sample items will be provided here for this scale or for the
scales below. [30] reported a Cronbach alpha of .91 for the
15-item version of the TI scale. [43] and [44] in their study
reported a Cronbach alpha of .895 for a 13-item version of
the scale. The reliability of the TIS-6 is however unknown.
Alienation Scale. Personal alienation (AL) was measured
with a five-item scale [9; 33]. The measure used in this
study was originally an 18-item measure, developed by
[33] to measure work alienation. [10] performed a principal
components analysis with a varimax rotation to assess the
factor structure of the alienation scale. [7] reported a
Cronbach Alpha of .80 for AL.
Helping Behaviour Scale (OCB). Organisational Citizen
Behaviour (OCB) was measured with a six-item Helping
Behaviour Scale [65]. [65] reported Cronbach Alphas for
the helping behaviour scale in a range from .95 to .88.
Maslach Burnout Inventory Human Services Survey
(MBI-HSS-20).
The MBI-HSS-22 was the original
instrument used by researchers to measure burnout in
occupations providing human services [46].
This
measurement was later adapted to measure job burnout of
occupations not focused on providing human services. The
MBI-General Services (MBI-GS) burnout dimensions are
slightly broader conceptualised and renamed as exhaustion,
cynicism and reduced professional efficacy [46]. This
instrument is copyright protected. To measure job burnout
for this study, the Maslach Burnout Inventory - Human
Services Survey (MBI-HSS-20) was used [27; 45; 66] with
three sub constructs labelled emotional exhaustion (eight
items), depersonalisation (five items), and reduced personal
accomplishment (seven items). Items 12 and 16 were
excluded due to consistent findings of cross-loadings in
previous studies [27]. A Spanish study [27] yielded
Cronbach alphas of .85 for emotional exhaustion, of .58 for
depersonalisation and of .71 for reduced personal
accomplishment.
Utrecht Work Engagement Scale (UWES). The UWES-17,
the instrument used in most work engagement (WE)
research, was used. The 17-item instrument is composed of
three sub constructs namely vigour (six items), dedication
(five items) and absorption (six items) [61]. Reliability
analysis of the UWES [60] yielded Cronbach alphas for

four samples ranging from .82 to .75 for vigour, from .90 to
.88 for dedication, and from .77 to .70 for absorption.
Task Performance Scale.
Task performance (PER)
assessment by supervisors was measured with a nine-item
adapted scale [54; 68]. [54] reported a Cronbach alpha of
.91 for PER.
Work-based Identity Scale. Although previous attempts
were made to measure work-based identity (WI) [6; 16; 17;
18] no suitable measurement was found that complied with
the requirements of the theoretical definition of WI in this
study. To measure the facets of WI (selected for this
study), namely work-role centrality, person organisationfit, organisational, job, career, occupation and work
involvement, different scale items were sourced and
adapted to fit the 7-point intensity response scale anchored
at extreme poles [6; 37; 40; 42; 51; 57]. The initial scale
consisted of 36 items that were compiled from these
abovementioned authors scales. [14] reported a Cronbach
alpha of .95 for the 28 items finally included in the WI
scale.
Research procedure
The Universal Resource Locator (URL) of the online
survey was initially sent by e-mail to 50 employees,
randomly selected from the target population for a pilot
study. No problems were identified in completing the
survey, except that it took on average about 75 minutes to
complete. Two doctoral studies were covered in the same
survey. The URL was then sent via e-mail to potential
participants requesting their participation. The successful
completion of the questionnaire by the participant activated
the last survey instrument, in-role job performance, to be
completed by the supervisor. Weekly e-mail reminders
were sent out to the target population asking and reminding
them to participate. Participation was voluntary and no
incentives were provided to enhance participation.
Statistical analysis
The Statistical Consultation Service (Statcon) of the
University of Johannesburg used the SPSS programme to
conduct the statistical analysis on the data in two phases.
The first phase consisted of descriptive statistical analyses,
the factor and reliability analyses, and the calculation of
correlations between all the variables. The second phase
consisted of inferential statistical analyses, namely
independent sample t-tests [50].
RESULTS
A factor and reliability analysis was conducted on the TIS6. A single factor was extracted with an overall reliability
of = .80. Item loadings ranged between .36 and .81 on
the single extracted factor and item reliabilities ranged
between .73 and .81. These findings confirmed the
factorial validity as well as the reliability of the TIS-6.
The data profiles of the 84 employees who resigned from
the ICT company over the four-month period after the
survey was conducted, were compared with the data
profiles of the 88 employees drawn randomly from the
sample (n = 2,429) who stayed with the company.
Independent sample t-tests were conducted to compare the
different variable scores of the stayers and leavers. The
following analyses provide evidence that the turnover
intention score can be used as a proxy for actual labour
turnover. The guidelines of Cohen (1988, pp. 284 - 287)

were followed to calculate the effect size for independentsample t-tests, expressed as partial Eta-squared where .01;
.06; or .14 respectively indicates small, medium, or large
effect.
There was a significant difference in the turnover intention
scores of leavers (M = 5.14, SD = 1.26), compared to
stayers (M = 4.13, SD = 1.28; t(170) = 5.21, p < .01 (twotailed)). The difference in the means (mean difference =
1.01, 95% CI: 0.63 to 1.39) has a large effect (p2 = 0.14).
This finding supports the predictive validity of the TIS-6.
There was a significant difference in the work-based
identity scores of leavers (M = 4.16, SD = 1.22) and stayers
(M = 4.96, SD = 0.92; t(170) = -4.84, p < .01 (two-tailed)).
The difference in the means (mean difference = -0.79, 95%
CI: -1.12 to -0.47) has a moderate effect (p2 = 0.12).
There was a significant difference in the personal
alienation scores of leavers (M = 4.85, SD = 1.40) and
stayers (M = 4.12, SD = 1.37; t(170) = 3.44, p < .01 (twotailed)). The difference in the means (mean difference =
0.73, 95% CI: 0.31 to 1.14) has a moderate effect (p2 =
0.07).
There was a significant difference in the reduced personal
accomplishment (BO1) scores of leavers (M = 3.92, SD =
0.98) and stayers (M = 4.26, SD = 1.06; t(170) = -2.18, p <
.05 (two-tailed)). The difference in the means (mean
difference = -0.34, 95% CI: -0.65 to -0.03) has a small
effect (p2 = 0.03).
There was a significant difference in the emotional
exhaustion (BO2) scores of leavers (M = 3.06, SD = 1.36)
and stayers (M = 2.19, SD = 1.42; t(170) = 4.07, p < .01
(two-tailed)).
The difference in the means (mean
difference = 0.86, 95% CI: 0.45 to 1.28) has a moderate
effect (p2 = 0.09).
There was a significant difference in the depersonalisation
(BO3) scores of leavers (M = 1.93, SD = 1.31) and stayers
(M = 1.48, SD = 1.23; t(170) = 2.32, p < .05 (two-tailed)).
The difference in the means (mean difference = 0.45, 95%
CI: 0.07 to 0.83) has a small effect (p2 = 0.03).
There was a significant difference in the vigour (WE1)
scores of those who resigned (M = 3.96, SD = 1.38) and
stayers (M = 4.73, SD = 1.00; t(151) = -4.24, p < .01 (twotailed)). The difference in the means (mean difference = 0.78, 95% CI: -1.14 to -0.41) has a moderate effect (p2 =
0.09).
There was a significant difference in the dedication (WE2)
scores of leavers (M = 3.60, SD = 1.78) and stayers (M =
4.60, SD = 1.21; t (170) = -4.32, p < .01 (two-tailed)). The
difference in the means (mean difference = -1.00, 95% CI: 1.46 to -0.54) has a moderate effect (p2 = 0.10).
There was a significant difference in the absorption (WE3)
scores of leavers (M = 3.86, SD = 1.43) and stayers (M =
4.59, SD = 0.97; t(170) = -3.94, p < .01 (two-tailed)). The
difference in the means (mean difference = -0.73, 95% CI: 1.10 to -0.36) has a moderate effect (p2 = 0.08). These
findings above support the discriminant validity of the TIS6.
There was no significant difference in the helping
behaviour scores of leavers (M = 5.27, SD = 1.13) and
stayers (M = 5.51, SD = 0.95; t(170) = -1.57, p > .05 (twotailed)). The difference in the means (mean difference = 0.25, 95% CI: -0.56 to 0.06) was non-significant (p2 =
non-significant).

There was no significant difference in the task performance

of leavers (M = 5.51, SD = 1.07) and stayers (M = 5.76, SD
= 0.93; t(154) = -1.57, p > .05 (two-tailed)). The difference
in the means (mean difference = -0.25, 95% CI: -0.57 to
0.07) was non-significant (p2 = non-significant).
The discussion of the results follows in the next section.
DISCUSSION
The research objectives of the study were to evaluate the
reliability, the factorial, predictive, and discriminant
validity of the TIS-6. This study will add to the unitary
validity and reliability information of the TIS-6 and
contribute towards establishing its credibility for future use
in the scientific community.
The findings of the study can be summarised as follows:
An exploratory factor analysis (EFA) established that the
TIS-6 is a one-dimensional construct, thereby confirming
the construct [more specifically the factorial validity (refer
to Allen & Yen, 1979)] of the scale. The scale reliabilities
with items deleted (ranging between .73 and .81) and the
overall reliability ( = .80) of the TIS-6 are acceptable.
In order to establish the predictive validity of the scale the
TIS-6 mean score differences for those who resigned were
compared to a randomly selected group who stayed. The
obtained mean score differences were significant and the
effect size was large. The predictive validity of the scale is
hereby established to be used as a proxy for actual
turnover.
In order to establish the discriminant validity of the scale,
independent sample t-tests were conducted to establish
whether the different variable mean scores of leavers and
stayers differed significantly. Significant mean score
differences were found in eight of the ten remaining
variables with effect sizes ranging between moderate and
small. The discriminant validity of the TIS-6 is confirmed.
These results show that the TIS-6 is a reliable and valid
measure to assess turnover intention and to validly predict
actual turnover behaviour as was suggested by [31] and
[49].
Turnover intention in this study relates to a number of other
variables such as personal alienation; the three dimensions
of work engagement (vigour, dedication and absorption);
the three dimensions of burnout (reduced personal
accomplishment, emotional exhaustion, depersonalisation)
and not only to the suggested chain of resources and
demands [7; 8; 22; 58; 60].
The TIS-6 can therefore be used in business and academic
research to validly and reliably assess turnover intention or
to predict actual turnover.
A limitation of the study is that the TIS-6 was only applied
in a single organisation and in a single country setting. A
strength of the study was however that it was conducted
across different culture groups in the South African
context. A suggestion for future research may be to
compare the scores of the TIS-6 across different culture
groups. There may be a possibility that culture groups
respond differently to antecedents leading to turnover
decisions.
REFERENCES
(The full reference list is available upon request)