Professional Documents
Culture Documents
Chapter 4: Analysis and Intepretation
Chapter 4: Analysis and Intepretation
Chapter Overview
The chapter begins with a discussion of the plan of analysis. In the initial part, the
details of the profile of respondents and responding organizations are presented.
The first part of the analysis deals with an estimation of response rate and the
testing of reliability and validity of the study scales. Thereafter, research
hypotheses are tested and the findings presented.
Data analysis begins with an account of the profile of respondents and the
responding organizations. This is followed by an estimation of response rate and
item completion rate. In the first part of analysis, the scales are tested for
unidimensionality, reliability and validity.
Exploratory Factor Analysis (EFA) was carried out to test if each of the scales was
a one-factor model. Indicator and scale reliability estimates were generated.
Further, convergent and discriminant validity was established vis-a-vis the study
scales. Thereafter, research hypotheses were tested by deploying Tests of
Differences such as Independent Sample T-test and One-way ANOVA . This is
followed by Test of Association (Chi-Square Test) on those variables that were
found to be having significant differences.
78
Profile of Respondents
The respondents of the study were categorized on the basis of their demographic
and professional profile.
Gender: In terms of gender, 68% of respondents were males and 32% were
female employees. Table 4.1 represents the profile of the respondents based on
gender.
Cumulative
Gender Frequency Percent
Percent
Male 340 68.0 68.0
Female 160 32.0 100.0
500 100.0
Total
Designation Cumulative
Frequency Percent
Level Percent
Junior 95 19.0 19.0
Middle 240 48.0 67.0
Senior 165 33.0 100.0
Total 500 100.0
Experience
in Present Cumulative
Frequency Percent
position Percent
(Years)
0 to 5 353 70.6 70.6
6 & above 147 29.4 100
Total 500 100.0
79
Total Experience: In terms of total experience, respondents were classified into 2
groups. As can be seen in Table 4.4, 44.6% were in a band of upto 10 years
while those having 11 years and above total experience constituted 55.4% of
respondents.
Total
Cumulative
Experience Frequency Percent
Percent
(Years)
0 to 10 223 44.6 44.6
11 & above 277 55.4 100.0
Total 500 100.0
Age: In terms of age, respondents were classified into 3 groups. The highest
frequency occurs around a band of 31 to 40 years (40.8%) and 41 to 60 years
(34.4%) of age, followed by 21 to 30 years (24.8%) as shown in Table 4.5:
Age Cumulative
Frequency Percent
(Years) Percent
21 to 30 124 24.8 24.8
31 to 40 204 40.8 65.6
41 to 60 172 34.4 100.0
Total 500
All the responding organizations belonged to non-profit sector as this was the
chosen sector for the present study. The responding organizations were further
classified on the basis of nationality & size.
80
Table 4.6: Responding Organizations- Nationality
Cumulative
Nationality Frequency Percent
Percent
Indian 235 47.0 47.0
Foreign 265 53.0 100.0
Total 500 100.0
Size: Responding organizations were classified into different sizes on the basis of
number of employees. It was observed that organizations in the non-profit sector
varied significantly in terms of size. There were a sizeable number of organizations
with less than 50 employees. Similarly, almost similar number of organizations
were big with more than 500 employees. Thus to maintain comparable
frequencies, organizations with 1 to 50 employees were considered small, those
between 51-100 were considered medium, those with 101-500 were considered
large and those with more than 501 employees were considered very large. The
break-up is presented in Table 4.7 below:
Number of Cumulative
Frequency Percent
Employees Percent
1 to 50 171 34.2 34.2
51 to 100 64 12.8 47.0
101 to 500 131 26.2 73.2
More than
134 26.8 100.0
500
Total 500 100.0
81
The present study was carried out largely using the internet based resources such as
e-mails and an online survey tool. Out of the 1200 employees in non-profit
organizations initially contacted by the researcher, 501 employees provided their
responses, i.e., a response rate of 41.75% was achieved. The response rate giving
an overall total of 501 responses can be considered to be high enough to generate
consistent statistical results.
Before proceeding with data analysis and testing of the study hypotheses, it was
deemed fit to assess the scales for reliability and validity. A thorough measurement
analysis on research instruments is essential for several reasons. First, it provides
confidence that the findings accurately reflect the proposed constructs. Second,
empirically validated scales can be used directly in other studies (Flynn et al.,
1994)
Measurement analysis of the study scales highlights the key issues of factor
analysis, reliability and validity (Hair et al., 2009; Malhotra & Dash, 2011). The
study scales were subject to all the above tests in order to be considered fit for
further analysis. Table 4.8 shows the list of variables/ measures used in the study.
82
Table 4.8: Variables/Measures Considered in the Study
83
Using the above constructs and the indicative items, the researcher proceeded with
measurement analysis of the study scales as well as final data analysis to test the
research hypotheses.
Factor analysis tests the unidimensionality of the measurement scales to refine the
variables which are not related (Henson & Roberts, 2006) and therefore, helps to
eliminate multiple, overlapping constructs in research. Exploratory Factor Analysis
(EFA) was carried out in the current study to see whether items loaded on their
hypothesized factors. Principal component analysis with varimax rotation was
deployed in the exploratory factor analysis performed as suggested by researchers
(Kahn, 2006; Sass & Schmitt, 2010).
Exploratory factor analysis describes factors that are there among a group of data
and its main objective is to unfold the hidden factors which explain the covariance
between the measured data (Kahn, 2006). Ahire et al. (1996) suggested that the
scale development consists of performing an EFA to identify major factors
according to item-factor loadings and refining the scales using Cronbach’s alpha.
EFA was carried out to check unidimensionality of each scale separately.
Dimensionality of the scale consists of testing that items load highly on a single
factor (Hair et. al., 2009). Factor analysis is seen as an essential precondition for
reliability and validity (Anderson & Gerbing, 1991). Henson and Roberts (2006),
Delgado-Ballester et al. (2003) and Fabrigar et. al (1999) suggested that Eigen
Values more than 1 is most often used criterion in factor analysis.
Williams et al. (2010) and Liu and Treagust (2005) suggested that before carrying
out EFA, it is important to assess the factorability of the overall set of variables
and individual variables using Kaiser-Meyer-Olkin (KMO) measure of sampling
adequacy and the overall significance of correlation matrix with Bartlett’s Tests of
Sphericity. Ang and Huan (2006) as well as Liu and Treagust (2005) suggest that
KMO and Bartlett’s tests reveal the appropriateness of the data for factor analysis.
KMO quantifies the degree of inter-correlations among the variables and thus, tests
84
appropriateness of factor analysis. KMO values must exceed 0.50 before
proceeding with the factor analysis (Hair et al., 2009; Malhotra & Dash, 2011).
The KMO values of all the scales used in the present study were found to be
acceptable, indicating that the data was suitable for factor analysis. Another
indicator is Bartlett’s Test of Sphericity which examines the presence of
correlations among the variables. It provides the statistical significance that the
correlation matrix has significant correlations among at least some of the variables.
Thus, a significant Bartlett’s Test of Sphericity is required (Hair et al., 2009;
Malhotra & Dash, 2011). Because p = .000 (its significance is less than 0.05) for
all scales, it indicates that one could proceed with factor analysis. The results of
KMO and Bartlett Test of Sphericity for all scales are given in Table 4.9.
Once values of KMO and Bartlett’s Test of Sphericity were found to be acceptable,
EFA was carried out on each scale individually.
HYG Scale: On the basis of Eigen value greater than 1 heuristic, only one
principal component was extracted that accounted for 50.668% of the total
variance. Thus, results of EFA revealed that the scale was unidimensional. The
results are given in Table 4.10A and Table 4.10B:
85
Table 4.10 B: HYG Scale-Component Matrix
Items Component
HYG-1 .760
HYG-2 .754
HYG-3 .797
HYG-4 .495
MOT Scale: Only one principal component was extracted in MOT scale too, that
accounted for 54.587 % of the total variance. Thus, results of EFA revealed that
the scale was unidimensional. The results are given in the Tables 4.11A and 4.11B.
86
Table 4.11 B: MOT Scale-Component Matrix
Items Component
MOT-1 .731
MOT-2 .719
MOT-3 .791
MOT-4 .800
MOT-5 .807
MOT-6 .740
MOT-7 .765
MOT-8 .515
MAT Scale: In case of MAT scale too, only one principal component was
extracted that accounted for 64.455 % of the total variance revealing that the scale
was unidimensional. The results are given in the Tables 4.12A and 4.12B.
Items Component
MAT-1 .732
MAT-2 .833
MAT-3 .839
87
COXT Scale: In the case of COXT scale, 3 principal components were extracted,
accounting for 35.920%, 12.645% and 11.122% of the total variance. Thus, results
of EFA revealed that the scale was not unidimensional. The results are given in
the Tables 4.13A & 4.13B.
Component
Items 1 2 3
COXT-1 .731 .041 -.294
COXT-2 .584 .561 -.049
COXT-3 .692 .244 -.271
COXT-4 .652 -.500 -.088
COXT-5 .545 -.243 .563
COXT-6 .637 -.258 -.391
COXT-7 .456 .622 .402
COXT-8 .550 -.339 .449
3 components extracted.
88
As can be seen from the tables above, 3 principal components were derived in the
COXT scale. Out of eight items, six loaded onto factor 1, while one item each
loaded on factor 2 and 3. As suggested by researchers, it is important that all items
should load highly (>0.3 or hopefully 0.4) on one factor and at least three items
should load on each factor (Bawa, 2004; Tansey et al., 2001). Since factor 2 and 3
had only one item loading onto them, they could not be considered as a one-factor
model. Moreover, despite cross-loading, the loading value of both these items on
factor 1 is more than 0.45 which is greater than the recommended minimum
threshold (0.3 or 0.4). Another rule is that all communalities should be greater than
0.5 or ideally greater than 0.6 for a scale to be considered as unidimensional
(Costello & Osborne, 2005; MacCallum et al., 2001). In the present case, all
communalities were well above 0.6. Thus, keeping the above rules in mind, it was
decided to retain the scale in its original form.
Further, Jöreskog and Sörbom (2002) and Lindquist et al. (2001) have stated that
when examining one-factor models, theoretical and interpretive considerations
should always be used as the primary consideration. Theoretical issues should also
be considered when determining the number of factors to be specified or retained
in the model (Garver & Mentzer, 1999). In some situations, the quantitative rule
may be overridden by judgmental considerations. For example, a variable with a
cross-loading in the 0.35 to 0.50 range might be assigned to a particular factor if
the variable has a theoretical relationship with other variables that load highly on
that factor (Costello & Osborne, 2005; Velicer & Fava, 1998). In the present study,
although some cross-loadings were there, theoretical considerations, apart from
statistical results, prompted that it was correct to take up COXT scale as a one-
factor model.
CONT Scale: On the basis of Eigen value greater than 1 rule, only one principal
component was extracted that accounted for 53.246 % of the total variance. Thus,
results of EFA testified that the scale was unidimensional. The results are given in
the Tables 4.14A and 4.14B.
89
Table 4.14 A: CONT Scale-Total Variance Explained
Items Component
CONT-1 .665
CONT-2 .696
CONT-3 .756
CONT-4 .743
CONT-5 .783
1 component extracted.
After the one-factor model for each study scale is ascertained, estimation of the
statistical reliability and validity becomes essential ( Cooper & Schindler, 2006;
Hair et al., 2009). Thus the scales were tested for reliability and validity.
4.6.1 Reliability
90
from 0 to 1 but should preferably be 0.5 or close to it (Schumacker & Lomax,
2004; Wu, 2005). In the present study, indicator reliability was more than 0.5 or
close to it in most cases. Table 4.15 reveals the indicator reliability for indicators in
each scale.
91
4.6.2 Validity
Once unidimensionality and reliability of the scales have been assessed, the next
step involved is assessing validity (Gerbing & Anderson, 1988; Hair et al., 2009).
Validity is the degree to which a scale or set of measures accurately represent the
concept of interest (Hair et al., 2009). Both convergent and discriminant validity of
the scale were assessed.
92
all the items did not load on a single factor, thereby indicating that discriminant
validity exists among the study scales. Tables 4.17 A & B present the results of
EFA.
93
Table 4.17 B : All Scales Rotated-Component Matrix
Component
Items 1 2 3 4 5 6
HYG-1 .347 -.060 .612 .003 -.110 -.345
HYG-2 .357 -.093 .600 -.283 .228 -.093
HYG-3 .435 .017 .650 .110 -.149 -.103
HYG-4 .278 -.064 .312 .182 .230 .280
MOT-1 .609 -.363 -.154 -.177 .265 .135
MOT-2 .647 -.269 .288 -.214 .031 .029
MOT-3 .688 -.352 .015 -.036 .093 -.082
MOT-4 .669 -.357 -.127 -.016 .144 .078
MOT-5 .657 -.426 -.174 .020 .050 .081
MOT-6 .644 -.362 .068 .106 -.247 .151
MOT-7 .676 -.345 -.092 .143 .018 .002
MOT-8 .451 -.215 .223 .402 -.383 -.154
MA-1 .626 -.268 .101 .057 .171 .021
MA-2 .531 -.435 -.348 .088 .036 .076
MA-3 .515 -.444 -.193 .210 -.033 -.071
COXT-1 .485 .413 .076 .035 .420 -.176
COXT-2 .272 .427 .091 .424 .360 .223
COXT-3 .427 .496 .000 .288 .240 -.142
COXT-4 .558 .352 -.020 -.320 .019 -.233
COXT-5 .364 .296 .473 -.285 -.017 .123
COXT-6 .580 .294 -.245 -.053 .123 -.188
COXT-7 .160 .457 .237 .446 .006 .466
COXT-8 .468 .328 .115 -.319 -.255 .292
COXT-9 .560 .435 -.241 .011 .121 -.181
COXT-10 .389 .301 -.089 .382 -.304 -.306
CONT-1 .473 .369 -.251 -.108 -.109 .023
CONT-2 .560 .236 .037 -.060 -.473 .218
CONT-3 .594 .367 -.194 -.255 -.014 .234
CONT-4 .613 .253 -.314 .002 -.035 -.125
CONT-5 .559 .249 -.283 -.092 -.305 .028
94
Table 4.18 : Discriminant Analysis - Correlation between Constructs
Independent samples T-test was carried out in order to test the hypotheses H01 to
H020. For hypotheses H021 to H035, One-way ANOVA was deployed since it
involved comparing three or more groups.
These tests were conducted with study variables on the basis of Gender (Males
and Females), Experience in Present Position (Low and High), Total Experience
(Low and High), and Country of Origin (Indian and Foreign) as described below:
95
On the Basis of Gender (Males and Females)
H04: There is no significant difference in the mean scores of Job Context Factors
(COXT) , between male and female employees.
96
H05: There is no significant difference in the mean scores of Job Content
Factors (CONT), between male and female employees.
Tables 4.19A and 4.19B present group statistics and results of t-test:
97
On the Basis of Experience in Present Position (Low & High)
Significant differences were not observed on the dimension HYG (t =-.537), p (or
sig 2-tailed) > 0.05 between employees having low level of experience in present
position (Mean=3.64, SD=0.745) and high level of experience in present position
(Mean=3.68, SD=0.862). Thus, the null hypothesis H06 was not rejected.
Significant differences were not observed on the dimension MOT (t = .929), p (or
sig 2-tailed) > 0.05 between employees having low level of experience in present
position (Mean=4.10, SD=0.666) and high level of experience in present position
(Mean=4.03, SD=0.743). Thus, the null hypothesis H07 was not rejected.
Significant differences were not observed on the dimension MAT (t = .048), p (or
sig 2-tailed) > 0.05 between employees having low level of experience in present
position (Mean=4.13, SD=0.794) and high level of experience in present position
(Mean=4.12, SD=0.754). Thus, the null hypothesis H08 was not rejected.
98
H010: There is no significant difference in the mean scores of Job Content
Factors (CONT) between low and high level of experience in present position.
Tables 4.20A and 4.20B present group statistics and results of t-test:
Std.
Construct Present Experience N Mean Std. Error Mean
Deviation
0 to 5 Years 353 3.64 .745 .040
HYG
6 Years and above 147 3.68 .862 .071
0 to 5 Years 353 4.10 .666 .035
MOT
6 Years and above 147 4.03 .743 .061
0 to 5 Years 353 4.13 .794 .042
MAT
6 Years and above 147 4.12 .754 .062
0 to 5 Years 353 3.53 .699 .037
COXT
6 Years and above 147 3.38 .618 .051
0 to 5 Years 353 3.64 .764 .041
CONT
6 Years and above 147 3.49 .815 .067
99
On the Basis of Total Experience (Low & high)
Significant differences were not observed on the dimension HYG (t =1.317), p (or
sig 2-tailed) > 0.05 between employees having low level of total experience
(Mean=3.70, SD=0.703) and high level of total experience (Mean=3.61,
SD=0.836). Thus, the null hypothesis H011 was not rejected.
Significant differences were not observed on the dimension MOT (t = .760), p (or
sig 2-tailed) > 0.05 between employees having low level of total experience
(Mean=4.11, SD=0.683) and high level of total experience (Mean=4.06,
SD=0.695). Thus, the null hypothesis H012 was not rejected.
100
H015: There is no significant difference in the mean scores of Job Content
Factors (CONT) between low and high level of total experience.
Tables 4.21A and 4.21B present group statistics and results of t-test:
101
On the Basis of Country of Origin (Indian & Foreign)
Significant differences were observed on the dimension HYG (t =-2.556), p (or sig
2-tailed) < 0.05 between employees of Indian organizations (Mean=3.56,
SD=0.818) and employees of foreign organizations (Mean=3.73, SD=0.737). Thus,
the null hypothesis H016 was rejected.
Significant differences were not observed on the dimension MOT (t =-0.503), p (or
sig 2-tailed) > 0.05 between employees of Indian organizations (Mean=4.06,
SD=0.688) and employees of foreign organizations (Mean=4.09, SD=0.709). Thus,
the null hypothesis H017 was not rejected.
Significant differences were not observed on the dimension MAT (t =-1.598), p (or
sig 2-tailed) > 0.05 between employees of Indian organizations (Mean=4.07,
SD=0.747) and employees of foreign organizations (Mean=4.18, SD=0.810).
Thus, the null hypothesis H018 was not rejected.
102
H020: There is no significant difference in the mean scores of Job Content
Factors (CONT) , between organizations based in India and based in other
countries.
Tables 4.22A and 4.22B present group statistics and results of t-test:
103
4.7.2 One Way ANOVA
These tests were conducted with study variables on the basis of Age (Various age
bands), Designation level (Junior, Middle & Senior) and Size (Number of
employees in the organization), as described below:
104
H025: There is no significant difference in the mean scores of Job Content
Factors (CONT) between employees in various age bands.
Tables 4.23A and 4.23B present ANOVA descriptives and results of ANOVA:
105
On the Basis of Designation Level (Junior, Middle and Senior level)
Tables 4.24A and 4.24B present ANOVA descriptives and results of ANOVA:
106
Table 4.24 A: ANOVA Descriptives for Designation Level
Sum of Mean
Construct Df F Sig.
Squares Square
Between Groups 4.578 2 2.289 3.803 .023
HYG Within Groups 299.184 497 .602
Total 303.762 499
Between Groups .313 2 .157 .329 .720
MOT Within Groups 236.812 497 .476
Total 237.125 499
Between Groups 1.202 2 .601 .983 .375
MAT Within Groups 303.913 497 .611
Total 305.115 499
Between Groups .812 2 .406 .879 .416
COXT Within Groups 229.370 497 .462
Total 230.182 499
Between Groups 1.782 2 .891 1.461 .233
CONT Within Groups 303.136 497 .610
Total 304.918 499
107
On the Basis of Size (Small, Medium, Large and Very large)
H031: There is no significant difference in the mean scores of Hygiene Factors
(HYG) between small, medium, large and very large organizations.
Significant differences were observed in the mean scores of small, medium, large
and very large organizations on HYG (p or sig <0.05). Thus, the null hypothesis
H031 was rejected.
Significant differences were not observed in the mean scores of small, medium,
large and very large organizations on MOT (p or sig >0.05). Thus, the null
hypothesis H032 was not rejected.
Significant differences were not observed in the mean scores of small, medium,
large and very large organizations on MAT (p or sig >0.05). Thus, the null
hypothesis H033 was not rejected.
Significant differences were not observed in the mean scores of small, medium,
large and very large organizations on COXT (p or sig >0.05). Thus, the null
hypothesis H034 was not rejected.
Significant differences were not observed in the mean scores of small, medium,
large and very large organizations on CONT (p or sig >0.05). Thus, the null
hypothesis H035 was not rejected.
Tables 4.25A and 4.25B present ANOVA Descriptives and results of ANOVA:
108
Table 4.25 A: ANOVA Descriptives for Size of Organization
Sum of Mean
Construct df F Sig.
Squares Square
Between Groups 9.328 3 3.109 5.238 .001
HYG Within Groups 294.434 496 .594
Total 303.762 499
Between Groups 1.237 3 .412 .867 .458
MOT Within Groups 235.888 496 .476
Total 237.125 499
Between Groups 2.877 3 .959 1.574 .195
MAT Within Groups 302.238 496 .609
Total 305.115 499
Between Groups 3.159 3 1.053 2.301 .076
COXT Within Groups 227.023 496 .458
Total 230.182 499
Between Groups .219 3 .073 .119 .949
CONT Within Groups 304.699 496 .614
Total 304.918 499
109
4.7.3 Summary of Tests of Differences
Note: Shaded boxes depict factors which are found to have significant differences
110
4.8 Tests of Association
Significant association does not exist between Nationality (i.e. Indian and Foreign
Organization) and perception on Hygiene factors as sig > .05. The results are shown
in table 4.27 A & B below.
111
Association between Hygiene Factors and Size of Organization
Significant association does not exist between Designation Level and perception on
Hygiene factors as sig > .05. The results are shown in table 4.29A & B below.
112
Table 4.29 B: Chi-Square Tests (Designation level)
Significant association does not exist between Gender (i.e. male and female) and
perception on Motivator factors as sig > .05. The results are shown in table 4.30 A
& B below.
Significant association does not exist between Gender (i.e. male and female) and
perception on Mission Attachment factors as sig > .05. The results are shown in
table 4.31 A & B below:
113
Table 4.31 A: Crosstab (Gender)
114
Association between Job Context Factors and Gender
Significant association does not exist between Gender (i.e. male and female) and
perception on Job Context factors as sig > .05. The results are shown in table 4.33A
& B below.
Significant association does not exist between Experience in Present position and
perception on Job Context factors as sig > .05. The results are shown in table 4.34
A & B below.
115
Table 4.34 B: Chi-Square Tests (Present Experience)
Significant association does not exist between Gender (i.e. male and female) and
perception on Job Content factors as sig > .05. The results are shown in table 4.35
A & B below.
Significant association does not exist between Nationality (i.e. Indian and Foreign
Organization) and perception on Job Content factors as sig > .05. The results are
shown in table 4.36 A & B below.
116
Table 4.36 A: Crosstab (Nationality)
Significant association does not exist between Experience in Present Position and
perception on Job Content factors as sig > .05. The results are shown in table 4.37
A & B below.
117
Association between Job Content Factors and Total Experience
Significant association does not exist between Total Experience and perception on
Job Content factors as sig > .05. The results are shown in table 4.38 A & B below.
Significant association does not exist between Age and perception on Job Content
factors as sig > .05. The results are shown in table 4.39 A & B below.
118