Research Approach on self-esteem and employee burnout as predictors of employee turnover
intention
The research on which this thesis is based was designed with the following characteristics:
A convergent parallel mixed methods design was used; it was both a qualitative and
quantitative study. As a quantitative study it measured different variables, as a qualitative
study it allowed for the subjective perspective of the different variables. Both quantitative
and qualitative data were treated with equal priority. Quantitative data is presented in chapter
fiver and qualitative data is presented in chapter six.
It was non experimental.
It was based on primary data collected through self-report scales, questionnaires and focus
group interviews.
The strength of this design is that it has relative potential to be generalised to larger populations
based on appropriate sampling design, high measurement reliability from the questionnaires/
scales selection or construction and high construct validity. Punch (2009) stated that researchers
using mixed methods need competencies in both quantitative and qualitative methods. Further
Punch (2009) postulated that the researcher needs to know how to complement the findings and
discussion of each method. The mixed methods design implemented in this study was used to:
converge the results from different data sources, increase and expand inquiry and to enhance
significant findings to give insight on employee turnover intentions, burnout and self-esteem.
However, it was complex to collect and analyse the qualitative and quantitative data at the same
time, and expertise was needed to apply both techniques. In addition, there was possibility of
sampling error, questionnaire error, high refusal and non-response rate, data capturing errors and
the inappropriate selection of statistical techniques. These issues are appropriately addressed
below.
In summary, this mixed methods study was used to measure the relationship between self-
esteem, employee burnout and employee turnover intention among professional counsellors in
Kenya. A convergent parallel mixed methods design was used. In this study, quantitative data
was used to test whether self-esteem and employee burnout predict employee turnover intention
among professional counsellors in Kenya. The qualitative data was collected to explore the
meanings and perceptions of counsellors concerning self-esteem, burnout and employee turnover
intention. The reason for collecting both quantitative and qualitative data is to provide a clear
understanding to the complex concepts of self-esteem, burnout and employee turnover intention.
Research Procedures
To achieve the stipulated research objectives, the following procedures were followed.
Permission to carry out the Study
Permission to carry out the study was obtained through the relevant structures, namely the
University of South Africa - Psychology Research Ethics Committee and the Kenya Counselling
and Psychological Association.
The participants were issued with an informed consent form asking them for their permission to
participate in the study. More details are given below (under ethical considerations) about steps
that were taken to safeguard participants.
Sampling Framework
The study utilised non-probability sampling. Convenience or opportunity sampling was used in
the context of a non-experimental research design. The sample focused on those respondents
willing to participate in the research; the disadvantage is that the results will not be exactly
representative of the wider population. However, as earlier mentioned in chapter one; the KCPA
has an estimated 2000 members and every year each member is required to renew their
subscription, the members who are up to date with their annual subscription are then classified as
active members.
As at April 2014, fully paid-up active members numbered 600, the average number of
counsellors attending the quarterly general meetings is 350.
Bias in the Sample
Given the voluntary nature of participation and the potential diversity, it was worthwhile
considering the possibility of bias manifesting in the sample. Consequently, the generalisations
made from research proposition were treated with caution. I am aware of the potential bias as
follows:
Interviewer bias: this could occur because the researcher participated in the distribution
of the questionnaire and when certain parts of the questionnaire were not clear
participants asked the researcher for clarification. The questionnaires were distributed
during the quarterly general meeting of the counsellors in May 2014.
Selection bias: this could occur because convenience sampling that was used in this study
only captured those participants who attended the general meetings of the KCPA.
In the same breath, because of convenience sampling, the research participants actually
volunteered, this posed a challenge of volunteer bias. The researcher acknowledges that
volunteers are not perfectly representative of the general population.
Quantitative Data Collection and Analysis
Measuring Instruments
The respondents filled in a questionnaire using the traditional pen and paper method. The
questionnaire had the following sections: demographic details, intention-to-stay questionnaire
(ISQ) - (Roodt, 2004); Maslach Burnout Inventory (Maslach & Jackson, 1981) and the
Rosenberg Self-esteem Scale (Rosenberg, 1965). The full questionnaire appears as appendix 1. A
few selected respondents participated in a focus group discussion after filling in the paper and
pen questionnaire.
Demographic Section
This section collected background information about the respondent; it included variables such
as age, gender, marital status, duration of working as a counsellor and current job status e.g.
contractual employment or self employment.
Intentions-to-stay Questionnaire
4.4.3.1 Rationale for Inclusion: The variable of turnover intention was measured by utilising
the intentions-to-stay questionnaire. This is an unpublished questionnaire by Roodt (2004). As
indicated in the problem statement, although the questionnaire deals with intentions to stay, the
theory and findings still hold valid for turnover intentions. Martin and Roodt (2008) reviewed a
number of studies and reported that most instruments measure turnover intentions on a small
number of items. They noted that some studies used a single item scale while others used more
than three items (Martin & Roodt 2008).
4.4.3.2 Composition of the Instrument: The questionnaire is made up of 15 items that are
measured on a five point Likert Scale. An example of an item is: “How often have you
considered leaving your job?” (‘Never,’ 1-low intensity to ‘always,’ 5-high intensity)
4.4.3.3 Reliability and Validity: The validity and reliability is relatively unknown for this
instrument. Jacobs (2005) reported a 0.913 Cronbach’s alpha co-efficient in a sample of
professional nurses in South Africa, and Martin (2007) reported a 0.895 Cronbach’s alpha co-
efficient in a sample of a South African tertiary institute, which indicates satisfactory internal
consistency.
Maslach Burnout Inventory
4.4.4.1 Rationale for Inclusion: The variable of burnout was measured by the Maslach Burnout
Inventory MBI (Maslach & Jackson, 1981). The questionnaire deals with burnout as a single
syndrome that has three definite components of emotional exhaustion, depersonalisation and
feeling of personal accomplishment. It is very widely used and has been validated for various
settings. The version used was the English version.
4.4.4.2 Composition of the Instrument: The MBI consists of 22 items distributed across the
three dimensions discussed above: (a) Emotional exhaustion - nine items, (b) depersonalisation -
five items, and (c) the feeling of personal accomplishment - eight items. The frequency of
burnout symptoms is rated on a seven point Likert scale ranging from “never” to “everyday”.
The MBI produces a separate score for each subscale. An example of an item on subscale of
emotional exhaustion: “I feel used up at the end of my workday”; an example on of an item on
subscale of depersonalisation “I have become more callous towards people since I took this job”;
and an example of an item on subscale of personal accomplishments is “I have accomplished
many worthwhile things in this job” (Maslach & Jackson, 1981).
4.4.4.3 Validity and Reliability: Schaufeli and Van Dierendonck (1995) reported satisfactory
internal consistency of the three subscales. They also report that it has convergent validity with
other burnout measures. Kim and Lee (2009), for a sample of social workers, reported a
Cronbach’s alpha of 0.91 for emotional exhaustion, 0.75 for depersonalisation and 0.79 for
personal accomplishment. Maslach and Jackson (1996) reported internal consistency at
Cronbach’s alpha of 0.90 for emotional exhaustion, 0.71 for personal accomplishment and 0.79
for depersonalisation.
They further demonstrated convergent validity and discriminant validity. Convergent validity
was demonstrated by correlating MBI scores with measures of various outcomes that had been
hypothesized to be related to burnout. Discriminant validity was obtained by distinguishing it
from measures of other psychological constructs that might be presumed to be confounded with
burnout. For example, there are important distinctions between burnout and depression.
Depression is a clinical syndrome, whereas burnout is description of a crisis in one’s relationship
with work especially the therapeutic relationship with service recipients (Maslach & Jackson,
1996).
Rosenberg Self-esteem (RSE) Scale
4.4.5.1 Rationale for Inclusion. The variable of self-esteem was measured by using the original
version of the Rosenberg Self-esteem scale. The Rosenberg Self-esteem scale measures global
self-esteem and has been widely used. This scale measures the individual’s positive or negative
attitude towards the self as a totality (Rosenberg, 1965). Vacha-Hasse, Kogan and Thompson
(2000) contended that even though the scale was developed 40 years ago, continued use of this
scale provides evidence of its reliability and validity.
4.4.5.2 Composition of the Scale: The Rosenberg Self-esteem scale is a 10 item Likert scale
with items answered on a four point scale from ‘strongly agree’ to ‘strongly disagree’. The
respondent is meant to read the statements and then indicate whether they strongly agree, agree,
disagree or strongly disagree. An example of an item of the scale: “On the whole, I am satisfied
with myself” (Rosenberg, 1965).
4.4.5.3 Reliability and Validity: The original sample for which the scale was developed
consisted of 5024 high school juniors and seniors from 10 randomly selected schools in New
York. Owens (2001) states that the scale has a Guttmann scale reliability coefficient of .92
among youth.
According to the Centre for Mental Health Services (1996), RSE has been useful for assessing
self-esteem in a variety of other groups, with test–retest correlations in the range of .82 to .88 and
Cronbach’s alphas for various samples in the range of .77 to .88 (Rosenberg, 1986). Lee et al.
(2010) in a sample of professional counsellors found internal consistency for the scores of all
items was .73, suggesting a moderately high degree of consistency across items. Silber and
Tippett (1965) obtained a test-retest reliability coefficient of .85 for 28 college students.
Rosenberg (1965) and Silber and Tippett (1965) tested the Rosenberg Self-esteem scale for
construct validity. They correlated the scores of the RSE with three other scales/measures of self-
esteem namely; Kelley repertory test, health self image questionnaire and the interviewers
ratings of self-esteem. The correlations were as follows respectively r =.67, r =.83, r =.56. These
convergent validities are significant for cross instrument correlations, which gives us evidence of
validity.
The Distribution of Questionnaires
Convenience sampling procedures was used to distribute the research questionnaires. The
questionnaires were distributed by hand during the quarterly general meeting of the Kenya
Counselling Association, which is normally held every three months. Not all counsellors attend
these meetings. The respondents were then expected to respond to them in their own time.
I expected to have an 80 percent response rate because the questionnaires were easy to fill and
took between six to 15 minutes to fill. The population sampled was motivated and anticipated the
benefits of the research. I offered to email scores and results to participants; and was available to
respond to any queries for clarification.
200 questionnaires were distributed, 180 questionnaires were returned of which completely and
correctly filled questionnaires were 162; this presents a response rate of 81 percent and an
accuracy rate of 90 percent. Incorrectly filled questionnaires included those that had un-
responded questions, and those whose responses were outside the range of the responses as
provided for in the Likert scales.
Statistical Analysis
The analysis consisted of two phases. The first phase entailed descriptive statistical analysis
describing simply the characteristics of the sample in relation to the variables. These included
frequencies, percentages, means, standard deviations and interrelations among the main
variables.
The second phase consisted of inferential testing. Phase two included the following statistical
tools:
Cronbach’s alpha was calculated to test the consistency with which the participants
responded to the items in each of the measures
Pearson product moment correlation: This was a test for measuring the relationship of the
variables by focusing on their interdependency. In this study, this was used to measure
the relationship age, duration of employment, academic qualification and job status and;
self-esteem, burnout, and turnover intention. It was also used to measure the relationship
between self-esteem, burnout and employee turnover intention.
Independent samples T test: this was used to compare the mean differences of the two
groups of a categorical variable, in this case the gender differences in relation to the
dependent variable namely: employee turnover intention and independent variables
namely: self-esteem and burnout. This is because one of the quantitative research
objectives is to measure the relationship between the background variables and the
dependent and independent variables.
Analysis of Variance (ANOVA): This was used to compare the effect of marital status,
academic qualification and employment status on turnover intention, burnout and self-
esteem scores.
Multiple regression analysis: Because the thesis had more than one independent variable
in the conceptual framework, multiple regression analysis was used to predict turnover
intention from levels of burnout and self-esteem scores.
Qualitative Data Collection and Analysis
Selection of Participants
Convenience or opportunity sampling was used to select participants for the group interview.
The group consisted of 23 participants. The total numbers of participants are the ones who
offered to participate in the group discussion. One group interview was held. The sample focused
on those respondents willing to participate in the research; among members of the Kenya
Counselling and Psychological Association. This association has an estimated 2000 members
and every year each member is required to renew their subscription, the members who are up to
date with their annual subscription are then classified as active members.
The participants were selected using the following criteria:
Above the age of 20
Be registered with KCPA as a Counsellor/Psychologist
Have participated in the quantitative part of the study
Be willing to participate in focus group discussion
Duration of employment and number of years of experience as Counsellor/Psychologist
was not part of the criteria; this was due to the desire to obtain diverse information from
the participants.
The participants were recruited at the annual general meeting of Kenya Counselling and
Psychological Association and were invited to participate through announcements at the meeting.
Data Collection
The chosen method of data collection for the qualitative part of the study was the focus group
discussion. The group met to exchange perspectives and opinions concerning the three variables.
I used group dynamics, which was a quick and cost effective method to enable participants to
freely share their perceptions concerning the variables as experienced. I allowed the participants
to respond in their own words to the themes of the variables. The disadvantage that I anticipated
was the introduction of bias. To counteract this, I ensured that group membership was
representative of the counselling population. The focus group was conducted in person i.e. the
participants all met in a conference room. The discussion was recorded on paper and taped to
enable accurate transcript and report.
Structure of the Focus Group Session
The focus group discussion started by participants introducing themselves, then some group
norms for the sessions were set; some of the norms set included: agreeing on confidentiality,
time management and language to be spoken, the agreed language was English more so because
it is the official language of instruction in schools and universities. After this, the objectives of
the focus group discussions were described and the participant’s expectations clarified. Then the
discussion took place. Some of the themes discussed included: a description of own experience
of self-esteem, own experience of burnout and own experience of employee turnover intention.
In addition participants were asked to try to explain in their own words what the descriptions
mean to them. Furthermore, they were asked to describe their own self-strategies for dealing with
low self-esteem, burnout and employee turnover intention. They were asked to provide
suggestions for individual counsellors, for counsellor supervisors and managers of counsellors on
how to deal with self-esteem issues, burnout and employee turnover intentions. Finally a
summary of the session was provided and participants were thanked for their contributions. The
focus group lasted for approximately 60 – 90 minutes.
Data Analysis
A transcript was made from the group discussion. The transcripts and reports were analysed
using thematic analysis. Using this method, I looked for patterns i.e. recurring words,
descriptions and phrases. Then the patterns were clustered into themes of related patterns.
Finally the patterns and themes were synthesized to answer the overall research questions.
Thematic Analysis
According to Holloway and Todres (2003) thematic analysis is seen as a foundational method for
qualitative analysis; by this they meant that it is a method of analysis that researchers ought to
learn because it equips them with core skills that are useful for conducting many other forms of
qualitative analysis. Braun and Clarke (2006) defined thematic analysis as a “method for
identifying, analysing and reporting patterns within data” (p. 80). Furthermore, Holloway and
Todres (2003) identify “thematizing meanings” as a shared generic skill within different methods
of qualitative analysis (p. 347).
Braun and Clarke (2006) postulated that this method is widely used without a clear agreement on
what it constitutes. They clarified that thematic analysis differs from other analytic methods that
seek to describe patterns from qualitative data (p. 80).
Consequently, Braun and Clarke (2006) identified a number of properties of thematic analysis
and these include:
Thematic analysis does not require detailed technical and theoretical knowledge of
approaches such as grounded theory and discourse analysis (p. 80).
It is not wed to any theoretical framework, so it can be used with different frameworks
and used to do different things. In this regard, it can be essentialist or realist in that “it
reports experiences, meanings and reality of participants”; it can be constructionist in that
it examines the ways in events, realities, meanings and experiences” (p. 80). Further, it
can be contextualist in that “it acknowledges the ways individuals make meaning of their
experience (p. 80).
It works to unravel the surface of reality and also to reflect reality (p. 80).
The Process of Thematic Analysis
This thesis implemented the process of thematic analysis as espoused by Braun and Clarke
(2006); this was as follows:
Step one: Becoming familiar with the data. This involved recording of all responses to
the discussion questions per question and reading and rereading the data, noting down
initial ideas (p.87).
Step two: Generating initial codes. This involved coding interesting features of the data
systematically from the responses of the participants and then collating data relevant to
each code (p.87).
Step three: Searching for themes. This was achieved by collating the codes into draft
themes and then collecting relevant responses in relation to each potential theme (p.87).
Step four: Reviewing themes. This was carried out by confirming whether the themes
work in relation to the coded extracts and the entire data set. (p.87).
Step five: Defining and naming themes. This involved an analysis for refining the
specifics of each theme hence generating clear definitions (p.87).
Step six: Producing the report. This was carried out by relating back the analysis to the
research question and literature, producing a report of the analysis (p.87).
Ethical Considerations
For the quantitative part of the study participant anonymity was maintained throughout the
questionnaire. The anonymity was intended to enhance the honesty of the responses given.
Participants’ identities remained anonymous by them not providing personal identifiable
information like name, name of workplace and residence. They were also required to complete
informed consent forms. Participants who wanted general feedback on the study outcomes were
requested to separately provide their names and addresses. It was also explained to the
respondents that the scales used in the questionnaire were used for study purposes
only and not for clinical diagnostic reasons.
For the qualitative part of the study the participants in the focus group discussion
agreed on maintaining confidentiality. They also agreed to share with the group
what they felt free to share. Each participant was given a participant number and
they referred to each other with those numbers as opposed to names so that the
recording and transcription could not capture their identities. After obtaining
permission from all the group members, the focus group was recorded.
