Model Specification and Assumption
A multiple regression model was used as it determines whether a group of variables
together predict a given dependent variable (Mertler & Reinhart, 2016). The model
separates each variable from the rest allowing each to have its coefficient describing its
relationship to the dependent variable. Saunders, Lewis and Thornhill (2017) state that
“multiple regressions are a statistical technique that can be used to explore the predictive
ability of independent variables on one dependent measure”. It is further argued that the
method is scientific as the data collected can be developed using systematic analysis
(Mugenda & Mugenda, 2012). The model analyses the relationship existing between a
dependent variable and one or several independent variables as used in the research study.
It is thus generally assumed that the correlation between the independent and dependent
variables are linear (Hair et al., 2009).
The research study was based on one multiple regression model to enabled the researcher to
investigate the relationship between various measures of psychological contract and
employee commitment as shown in model 1. Model 1 tested hypotheses 1-4. The research
study had its basis on one key multiple regression model, which facilitated the investigation
of the relationship between various measures of psychological contract and employee
commitment as well as the moderating effect of leadership styles on the relationship
between psychological contract and employee commitment.
Y=β0+β1CO +β2FO+β3EO+β4L+β5CO*L+β6FO*L+β7EO*L+𝜀……………………. model 1
Where: Y = dependent variable (Employee commitment).
β0 = Constant or intercept refers to the value of the independent value when the value of the
independent variable is zero.
β1-7= Regression coefficient for each independent variable.
CO= Content -oriented framework
FO= Feature-oriented framework
EO= Evaluation-oriented framework
L= Moderating Variable (Leadership)
𝜀= Stochastic or disturbance term or error term.
Assumptions for the Model
To conclude a population based on a regression analysis done on a sample, several
assumptions must be true (Berry, 1993). The study checked for the following two
assumptions of multiple linear regressions: Independence of errors- that assumes that for
any two observations the residual terms should be uncorrelated (or independent). This
assumption was tested with the Durbin-Watson test, which tests for serial correlations
between errors;
Multi-co linearity, which assumes that there should be no perfect linear relationship
between two or more of the predictors. So, the predictor variables should not correlate too
highly.
Research Instruments
A questionnaire was the main data collection tool used in the study. A questionnaire entails
a formalized list of questions useful in soliciting information from respondents. It was
mainly used to gather quantitative data from respondents. The nature of data collected
guided the selection of the questionnaire and this was carried out collected quickly and
efficiently as indicated by Mugenda and Mugenda (2012). It also facilitates data analysis,
both descriptive and inferential analyses. The questionnaire comprised of close-ended
questions based on a 5- point Likert scale. In the Likert scale, the arithmetic mean was used
to get the index. The results obtained from primary data analysis were tested for correctness
with the help of two parameters as validity and reliability.
Data Collection Procedures
The research instrument used for collecting primary data was a questionnaire, which is the
most widely used data collection method in evaluation research. The Questionnaires used
for the final data collection was close-ended.
Primary data was collected to facilitate understanding of the direct relationship between
employees and employers. It also ensured no third-party interference as with the case of
secondary data (Sekaran & Bougie, 2009). Data collection was done using structured
questionnaires and administration achieved by hand delivery. Although mail delivery was
fast and easy to collate, its limitations were that most of the county workers had no private
mailboxes or reliable access to emails, which could have lowered response time and rate.
Hand delivery increased response rate and thus time and efficiency during data collection.
Besides, it overcame challenges caused by the likelihood of some participants who could
ignore the questionnaires hence leading to low response rate and biases of the results which
could be difficult to trace.
Validity of the Research Instrument
Mugenda and Mugenda (2012) explain that instrument validity is attained when research
instruments are accurate in measuring that which they are designed. Sekaran (2008) noted
that the validity of the questionnaire data depends on a crucial way on the ability and
willingness of the respondents to furnish the information requested and more so concerns
the rate at which the test contents compare the domain of the content. The determination of
research validity involved posing a series of standardized questions through the use of a
questionnaire to the possible respondents in the sample.
Questionnaires are considered to lack validity for many reasons, as several respondents
may lie or give a response that is desired. Reliable measurements instruments are free of
random error. Therefore, validity is the degree to which an instrument or test measures
what it needs to measure, such as content, criterion, logical or construct validity. Validity
can also be used to denote the degree to which the scores from the test or instrument
measure its purpose. The questionnaire was sent to some experts like the supervisors and
other researchers to measure the content validity, and then it was modified based on the
suggestions offered by the experts. This test of validity ensured that it was consistent with
the objectives of the study and the research paradigm as suggested by Meyers and Welkom
(2016).
The researcher prepared the research instruments, such that they were able to measure the
target content. To further assess instrument content validity, the study computed the scale
content validity index of the questionnaire. Items in the instrument were rated on a scale of
1- 5 in an item-rating continuum ranging from strongly agree to strongly disagree as
espoused by Shrotryia and Dhanda (2019) where 5 was Strongly Agree, 4 was Agree, 3 was
Neutral, 2 was Disagree and 1 was Strongly Disagree. The average scale validity index was
later calculated using the ratio of a number of the ratings to the total number of items. An
index of 0.80 was accepted by the researcher as recommended by Shrotryia and Dhanda
(2019).
Construct validity was achieved by clustering questions around a few concepts and cross-
loading of items with multiple factors kept minimal. Content validity was achieved by
correlating the kinds of leadership styles provided by the Migori County Government and
the level of work commitment by the county employees. Face validity was achieved by
using the right tools, creating a table of specifications and handling the tools like
questionnaires in the right manner so as not to create complications that can reduce the
reliability of the instruments.
Reliability of the Research Instrument
Data reliability is a measure of internal consistency and average correlation. It is the extent
to which a test, measurement process or research instrument generates a common result on
repeated treatments or trials. Hooley (2008) states that reliability refers to the degree to
which a measure yields stable and error-free results. In other words, reliability ensures the
stableness of the measurement procedure. The process was treated as reliable since the
measurement device stably assigned a similar score to objects or individuals with common
values.
Reliability implies that scores of an instrument are stable and consistent thus ascertaining
that the research instrument can be relied upon. A reliable score is reproducible and
consistent. Research instrument reliability was measured using Cronbach’s Alpha
coefficient; which was considered appropriate since the questionnaire was largely a
psychometric instrument measuring the perceptions, opinions, and attitudes of the
respondents. According to Hardy and Bryman (2009), Cronbach’s alpha index of 0.7 or
higher is recommended in judging the reliability of a research instrument. Instrument
reliability index of 0.789, 0.791, 0.784 and 0.794 was achieved for leadership styles,
content- oriented, feature-oriented, and evaluation-oriented respectively; implying a high
level of internal consistency for the scale with the sample data. The findings of reliability
tests for each variable were as summarized in Table 3.1.
Table 3.1: Reliability Statistics
Variable Cronbach’s alpha No. of items
Content-oriented framework 0.791 5
Feature-oriented framework 0.784 5
Evaluation-oriented framework 0.794 4
Leadership style 0.789 4
Source: Researcher, 2022
Research Design
The study adopted an explanatory design. This design is explanatory as it “aims to establish
the relationship between the independent and dependent variable, describing how they exist
with each other while analysing the impact of individual variables on the dependent
variable, making it fall under the category of explanatory research” (Deepthi & Baral,
2013; Agarwal, 2011). The quantitative method was adopted for the current research
while the hypotheses
generated for this research were tested using inferential statistics to generalize the results. It
was not feasible to undertake qualitative research since it was not founded on the
exploration of the topic by an understanding of the individual participant’s perception.
Target Population, Sampling Techniques and Sample Size
This section explores the target population, sampling techniques and sample size.
Target Population
Mugenda and Mugenda (2012) define a population as a well-defined set of people, services,
elements and events, a group of things, a household that are under investigation Cooper and
Schindler (2013) define primary data as information collected from literature, the internet,
journals, databases, books, journals and report. Primary data can be both qualitative and
quantitative. A population is the total collection of elements on which research is done. A
target population thus is that population which the researcher wants to generalize results
(Mugenda & Mugenda, 2012). The population of this study comprised employees of the
Migori County government, both permanent and casual, whose number stood at 945
according to the County Human Resource record. Of this total, 54% were female
employees while the other percentage, 46% were male employees (Ogoye, 2013). The
summary is given in Table 3.2.
Table 3.2: Target Population
Cluster Awendo Uriri Rongo Suna East Total
Top level cadre 20 15 10 20 65
Middle level 50 70 100 40 260
Lower level 100 200 150 170 620
Total 170 285 260 230 945
Source: Migori County HR Department (2022)
Sampling Techniques
The research applied a cluster sampling technique to select employees from the county
government. Management levels were thus arrived at using the cluster sampling technique.
The clusters included middle level and lower levels management divided into homogeneous
yet internally heterogeneous groupings and thereafter selecting random groups (clusters)
with systematic random sampling for data analysis.
Sample Size
According to Mugenda and Mugenda (2012), sampling refers to the act, process or
technique of selecting a suitable representative part of the population for determining
parameters to assume the whole. In other words, a sample is part of the whole population
on which the study is done. A sample is the segment of the population that is selected for
investigation.
Anderson and Maxwell (2016) suggested another simplified formula for the calculation of
sample size from a population which is an alternative to Cochran’s formula. According to
him, for a 95% confidence level and, the size of the sample should be;
𝑁
𝑛 = 1 + (𝑒)2
Where N is the population size and e is the level of precision
This formula is used for a population, in which N =945 with ±5% precision. Assuming a
95% confidence level and p =0.05, we get the sample size as
𝑁
𝑛 = 1 + 𝑁(𝑒)2
𝑁
945
945 945
𝑛 = 1 + 𝑁(𝑒)2 = 1 +
945(0.05)2 =
=
1 + 2.3625 = 281
3.3625
Mugenda and Mugenda (2012) note that at least 20% of the total population is
sufficient in descriptive studies. For this study, a sample of 30% of the total
population was thus considered sufficient and was used; therefore 281
respondents comprised the sample size for the study as shown in Table 3.3.
Table 3.3 Sample Size
Category Population
frequency
Sample size Rongo Awendo Uriri Suna-East
Top level Cadre 65 18 3 5 4 6
Middle level Cadre 260 78 30 15 21 12
Lower-level Cadre 620 185 44 30 60 51
Total 945 281 77 50 85 69
Source: Author (2022Pilot Test
This is a small-scale preliminary study conducted to assess the feasibility of the
main study and helps in assessing the validity and reliability of the research
instruments to be used, as well as the process considerations while administering
the questionnaires. Pilot testing enabled the researcher to identify shortcomings
that could occur, inadequacies of the research process, including any other
challenges and issues that were likely to occur while conducting the research
study. For determination of the research questionnaire reliability, the researcher
conducted a pilot study with 10% of the total sample to whom the instrument
was administered, representing 28 non-sampled respondents from three Sub-
counties not selected for the study; Nyatike, Kuria East and Suna West. The
number of respondents for a pilot study should be between 9% and 10% of the
sample of the study (Hardy & Bryman, 2009).
The pilot study addressed broad concerns which included availability of the
respondents, how their daily work schedules were respected, understanding of
the items in the instrument, acceptability of the method used in data collection,
willingness of the respondents to answer questions and how much time needed to
administer the questionnaire.
Data Presentation and Analysis
Data Analysis
Data analysis refers to examining what had been collected in a study and making
deductions and inferences (Donald & Deno, 2016). It involves uncovering
underlying structures, extracting important variables, detecting any anomalies
and testing any underlying assumptions. This study analysed data descriptively
as well as using inferential statistics. The inferential statistics included Pearson
correlation, T-test, and multi-linear regression analysis.
The study used descriptive statistics (frequencies, frequency percentages and
mode) to identify the aggregate patterns in the study variables for a total of 281
observations. The views of the respondents were captured in a 5-point Likert
scale to show their(respondents) level of agreement or disagreement to the
statement where (5= Strongly Agree (SA), 4=Agree (A), 3=Neutral (N),
2=Disagree (D) and 1=Strongly Disagree (SD). The frequencies and Likert
scales were thereafter used to compute the Relative Importance Index values of
each statement in the independent variables using the formula;
5𝑛5+4𝑛4+3𝑛
3+2𝑛2+1𝑛1
𝐴∗𝑁
Whereby: n1=number of respondents for
strongly disagree, n2=number of respondents for
disagree
n3=number of respondents for
neutral n4=number of respondents
for agree n5=number of
respondents for strongly agree
A=(Highest weight)=5 and N=(Total number of participants)=281
Data Presentation
The results were presented using frequency tables.
Measurements of Variables
The dependent variable was employee commitment which was measured by
normative. Independent Variable indicators were measured using an interval
scale.Table 3.4 Measurements of Variables
Variable Measurement Scale Question
number
Content-
oriented
Job content and terms are conducive
The management style plays a role in accommodating staff needs
There is an observed internal relation in the department
The policy development and implementation are inclusive
Career development of staff has increased the staff commitment
Interval
2.0
Feature-oriented There is job stability in the department
The organization structure is both implicit/explicit
Trade Union presence plays a control role
Interval 2.1
Evaluation-
oriented
Change in the department is professionally managed
Needs and expectations are considered
Violations in the department affect employee obligations
Breach of trust in the department impacts employees
Interval
2.2
Employee
commitment
(Normative
commitment)
The employees are satisfied by the way the employer invests in them
The employees would like to pay back the organization for all the
investments made in them in terms of training
The employee is positive about the organization’s future
in the workplace
The employee would want to reciprocate to the employer for the fair
treatment given to them
Interval 1.2
Leadership
Styles The employee enjoys the leadership style in place
The leadership style in the organization is professional
The employee has a chance to raise concerns to the leader
The leadership style in the organization considers employee’s needs
Interval 1.3
Source: Rousseau and Tijoriwala (1999)
Hypothesis Testing
Inferential statistics were used to test the study hypotheses. The study used a
95% confidence interval. A 95% confidence interval indicates a significance
level of 0.05. This implies that for an independent variable to have a significant
consequence on the dependent variable, the p-value ought to be below the
significance level of (0.05). This model treated the psychological contract as the
independent variable and employee commitment as the
dependent variable. To arrive at empirical conclusions tests of various
hypotheses were done. Table 3.5 indicates the summary of the research
hypothesis, decision rule and the interpretation of the expected results.
Table Hypothesis Testing
Hypothesis Statement Type of Analysis Decision Rule and Interpretation
H01: There is no significant effect of the content-
oriented framework on employee commitment of
employees in the Migori County Government.
Regression-
T- Test
H01: ß1=0
Ha: ß1 ≠0
Reject H01 if P-value is ≤0.05 at
α=0.05 otherwise fail to reject
H01
H02: There is no significant effect of the feature-
oriented framework on employee commitment of
employees in the Migori County Government.
Regression-
T- Test
H02: ß2=0
Ha: ß2 ≠0
RejectH02 if p-value ≤0.05
at α=0.05 otherwise, fail to reject
H02
H03: There is no significant effect of the
evaluation-oriented framework on employee
commitment of employees in the Migori County
Government.
Regression-
T- Test
Ho: ß3=0
Ha: ß3 ≠0
Reject H03 if P-value is ≤0.05 at
α=0.05 otherwise fail to reject
H03
H04: Leadership style has no statistically
significant moderating effect on the relationship
between psychological contract and employee
commitment.
Regression-
T- Test
H04: β5, β6, β7=0
Ha4: β5, β6, β7≠0
Reject H04 if p-value≤0.05 at
α=0.05 otherwise fail to reject
H04.
Source: Researcher, 2022
Ethical Considerations
The ethical standards for this research were maintained by taking responsibility for the value of self-
esteem and self-respect of the respondents. Statement of privacy and confidentiality was attached
alongside the questionnaire for the participants to read, and accept or agree by signing. Responses
remained private and confidential and for academic use only. The identity of the respondents was
concealed and anonymity was maintained. The researcher informed the respondents about privacy and
confidentiality, the value of the study. The study was sensitive to human dignity. After a successful
proposal presentation, the researcher sought an introductory letter from the School of Post Graduate
studies Rongo University, which facilitated authorization by the national commission for science and
technology and innovation (NACOSTI). The research assistants identified for the study were trained in
the research tool to ensure accuracy and efficiency. Permission was obtained from relevant authorities
before the start of the research. There was no discrimination of the respondents and foul language was
avoided during the study.
