This chapter outlines in detail the methodological steps and procedures for the proposed study. It contains the research framework, hypotheses, research procedures, research design, data collection, sample, method of analysis, and measurements, followed by validity and reliability.
Research Framework
The research framework is based on research questions stated in chapter 1, as well as hypotheses formulated after the review of literature. The purpose of this research is to compare the relative influences of perceived customer appreciation, work motivation and perceived supervisor support on employee job satisfaction.
Hypotheses
General hypotheses are derived from the research framework and stated as follows:
Figure 3.1. Research framework.
Hypothesis 1: Perceived customer feedback positively influences employee job satisfaction.
Hypothesis 2: Perceived supervisor support positively influences employee job satisfaction.
Hypothesis 3: Perceived customer feedback positively influences work motivation.
Hypothesis 4: Perceived supervisor support positively influences work motivation.
Hypothesis 5: Work motivation positively influences employee job satisfaction.
Research Procedure
The research procedure describes the sequence in which the researcher took to carry out this study. Figure 3.2 illustrates the steps that were taken to formulate and complete this study. In addition, each stage is described below.
The researcher first read through journal articles in order to find a topic of interest. This introduced the researcher to many different research questions, findings, and implications for possible future researches. From here, the researcher formulated broad questions before narrowing down to a specific topic of interest. After narrowing down research questions to the topic of interest, the researcher developed multiple research frameworks before settling on one to appropriately illustrate the relationships among variables chosen for this study.
More literature was then reviewed to describe and clarify the relationships between and among variables in the research framework and proposed study. Throughout the first 5 stages of this research, the researcher continued to develop hypotheses until after the framework was finalized.
The researcher chose a quantitative research approach while it was identified as the most appropriate method for collecting data for this study. In addition, appropriate scales of measurement for quantitative data collection were found and utilized. Scales of measurement for this research are from developed questionnaires and were adopted or adapted in order to measure the variables of this study.
The researcher conducted a pilot study to test the feasibility of the full-scale study. After collecting and analyzing results from the pilot study, the researcher reviewed the instrument, and made adjustments to modify the instrument by reordering the questionnaire.
Surveys were administered in questionnaire form. The questionnaires were distributed through personal distribution, referrals and online tactics. Data collected was analyzed through various statistical tools including, but not limited to descriptive statistics, factor analysis, alpha coefficient tests, correlation analysis, regression, and structural equation modeling. Lastly, analysis of the data allowed the researcher to conclude findings that corresponded with the study. The researcher then provided suggestions and implications for future studies.
Figure 3.2. Research procedure.
Conclusion and Suggestions Data Analysis
Data Collection Review of Instrument
Pilot Study
Adoption of Instruments Research Design Development of Hypothses Collection of Relative Literature
Design of Framework Selection of Topic of Interest IdentiCication of Research Questions
Review of Literature
Research Design
The research design for this study took a quantitative approach in order to test the five hypotheses. Through a cross-sectional and causal research approach, this study conducted a surveyon service industry personnel and collected quantitative data through questionnaire instruments. Once data was collected, the study utilized inferential statistics to test the hypotheses.
Data Collection
Since this study proposed to adopt a quantitative research approach, it employed data-gathering methods in survey form by distribution of questionnaires. The questionnaires were personally distributed and handed out through referrals as well as was accessible online. Each questionnaire recipient received a cover letter, followed by items from the measurements of this research’s four variables. The questionnaire adopted, adapted, and utilized items from developed instruments. Detailed descriptions of each measurement scale are described below under the appropriate headings. The questionnaire consisted of 62 items to measure the four variables, as well as 12 questions of demographic or screening nature. Scores were rated on either 5-point or 7-point Likert scales and totals averaged out to find mean scores.
Sample
This study proposed to employ a cross-sectional research design by collecting data from individuals employed in any area within the service industry. The population of this study consisted of individuals working under the supervision of supervisors in the service industry who have face-to-face interactions with customers. Since there was no available sampling frame for this population, convenience sampling method was used. Survey respondents were from for-profit, nonprofit, and public sector organizations based in and who worked within the state of Hawaii. The cover letter provided for the questionnaire guaranteed anonymity of respondent’s answers.
Sample Profile
For the full-scale study, 328 questionnaires were returned (70 online respondents, 258 hardcopy respondents). A total of 313 questionnaires were useable for this study. Eight hardcopies were excluded due to missing sections and seven online responses did not pass all screening questions; namely having face-to-face interaction with customer or clients and currently working in Hawaii. After reviewing the 313 questionnaires some of the following demographics were revealed. A majority of respondents at a rate of 74.12% reported as working in Honolulu, the capital of Hawaii. Responses came primarily from females at a rate of 60.70%. The majority of respondents reported as being between the ages of 26 to 35. Most respondents reported working for their current employers between 1 and 5 years. A rate of 70.62% reported to work in the business sector, while 13.73% and 15.66% stated as working in the public and nonprofit sectors, respectively. Reported levels of education completed are as follows: 53.67% high school; 7.35% certificate; 10.23% associates; 23.64% bachelor’s;
4.47% master’s; 0.64% other. A full summary of the sample’s characteristics is depicted in Table 3.1.
Table 3.1.
Descriptive Statistics on Sample Characteristics
Item Frequency Percentage Item Frequency Percentage
1. City of
Table 3.1. (continued)
The data collected from the sample was analyzed through statistical software IBM SPSS 21 and SPSS Amos using statistical analysis methods indicated in the following sections.
Descriptive Statistics
Since this study did not employ a random sampling method, the purpose of descriptive statistics mainly helped to calculate important features of the data such as sample mean, median, mode, standard deviation, demographics, and sample distribution. Descriptive statistics was beneficial towards this study while it summarized, described and characterized the prominent features of the quantitative data collected.
Factor Analysis
This study conducted exploratory factor analysis (EFA) in SPSS and confirmatory factor analysis (CFA) in AMOS to establish construct validity of the measurements. EFA was used to distinguish properties based on the data collected. EFA also maximized the amount of variance explained. Harmon’s single factor test is observable in EFA, which tests for common method variance (Podsakoff & Organ, 1986). While this research proposed to collect data through self-report questionnaires, there was no doubt this study would run into common method variance (CMV). CMV occurs when two or more measures come from the same source. Recommendation to minimize CMV include scale reordering, while keeping items within the same variable domain, as well as varying measurement scales.
In contrast to EFA, CFA is a statistical tool often used to observe relationships among latent variables (Jackson, Gillaspy Jr., & Purc-Stephenson, 2009) and evaluates a priori the study’s hypotheses. In order to utilize CFA, the researcher is required in advance to hypothesize a number of factors, and identify whether or not those factors are correlated.
Therefore, CFA tests whether the data collected fits the theorized measurement model of a variable.
Analysis of Variance (ANOVA)
ANOVA is a statistical tool that tests whether or not the means of several groups are equal. It’s purpose is similar to t-test analysis, however, it can be used to test and compare data of more than two groups. ANOVA was used to compare the means of different working
sector groups within the service industry among this study’s variables (customer feedback, perceived supervisor support, work motivation, and employee job satisfaction).
Correlation Analysis
This study used Pearson’s correlation analysis to test the relationship between variables.
This analysis method results in correlation coefficients, thus allowing the observation of when a variable changes, whether or not another variable will change. In addition, this analysis allows for the observation of strengths and significances of relationships between variables and selected demographic features.
Structural Equation Modeling (SEM)
Structural equation modeling (SEM) is a statistical procedure used to test and estimate causal relationships (Jackson et al., 2009). SEM technique uses both statistical data and causal assumptions. Despite the intricate amount of analysis output yielded in SEM, there is no set rule on what figures should be reported.
The data collected for this study was analyzed using AMOS (Analysis of Moment Structures), which is a type of SEM, in SPSS software. The distinctiveness of AMOS is that it automatically incorporates estimation of variances for all independent factors. AMOS also allows for creating path diagrams and CFA (Byrne, 2013). In addition, AMOS uses maximum likelihood estimation as its default feature (Bryne, 2001).
In SEM, confirmatory factor analysis and path analysis were conducted, producing results that answered several research questions as well as hypotheses. Each variable was independently analyzed in SPSS AMOS for confirmatory factor analyses, then path analysis was run to test the study’s hypotheses. Outputs of highest interest for this study included X2/df, RMR, GFI, AGFI, RMSEA, AVE and CR to examine measurement models’ goodness of fit.
X2/df refers to Chi-square divided by degrees of freedom, which is also known as the relative chi-square. It is a guide showing what the fit of data to the model is when dropping
pathways. The root mean square residual (RMR) is an indicator of how much estimated variances and covariances are different from the observed ones. Goodness of fit (GFI) is a statistic that analyzes the proportion of variance that is accounted for by the projected population covariance. Adjusted goodness of fit (AGFI), is an index of adjusted GFI parameter. Root mean square error or approximation (RMSEA) compares lack of fit to the saturated model (Hooper, Coughlan, & Mullen, 2008). Table 3.2 is a summary of model fit indices of X2/df, RMR, GFI, AGFI, and RMSEA. By meeting all cutoff criteria, this indicates goodness of fit.
Table 3.2.
Index of Model Fits
Good fit Acceptable fit Author’s notes
X2/df 2-5 <5
Note. Summary based on Hooper, Coughlan, and Mullen (2008) (top rows) and Schermelleh-Engel, Moosbrugger, and Müller (2003) (bottom rows).
Reliability of constructs in measurement and structural models are represented by composite reliability (CR) (Bacon, Sauer, & Young, 1995). As for average variance extracted (AVE), this is the average sum of a construct’s squared standardized loadings.
Henseler, Ringle and Sarstedt (2015) state AVE signifies the “average amount of variance that a construct explains in its indicator variables relative to the overall variance of its indicators.” Following Fornell and Larcker’s (1981) criteria, acceptable CR is above .80 and AVE is above .50. However, when factor loadings are above .707 for individual items within a construct, this is also considered adequate reliability (Khosrow-Pour, 2006).
Measurements
The survey questionnaire used the following instruments of measurement for each variable. The next section details each instrument of measurement that was adopted or adapted for this study. Permission for use of four instruments was obtained. Authors of the work motivation measurement (Schaufeli & Bakker, 2003) have indicated the scale may be used for free for noncommercial scientific research. Below are descriptions of each measurement and are written in the order presented in the full-scale study survey.
Employee Job Satisfaction (EJS)
To measure the dependent variable this study adopted Weiss et al.’s (1967) Minnesota Satisfaction Questionnaire (MSQ) short form, which included 20 items. After correspondence with Vocational Psychology Research of University of Minnesota, permission was granted to use the 1977 MSQ short form version upon payment of Royalty fees. Items statements of the MSQ asked respondents to rate their level of satisfaction with their current job. Example statements are “The chance to do things for other people” and “The chance for advancement on this job.” Items were scored on a 5-point Likert scale (1 = Very Dissatisfied; 5 = Very Satisfied). Cronbach’s alphas from previous researches using this measurement reported
ranges between .85 and .91 (Fields, 2002).
Work Motivation (WM)
Work motivation was measured using Schaufeli and Bakker’s (2003) “Utrecht Work Engagement Scale” (UWES) (Schaufeli, Bakker, & Salanova, 2006). The scale reported Cronbach’s alphas ranging from .70 to .90 in past studies. Similar to Bakker and Demerouti’s (2008) study that acknowledges motivation as an antecedent of engagement, this study also adopted the same perspective. Recognizing motivation as a predictor of engagement, by measuring engagement, results ought to reflect levels of motivation. The measurement contains 17 self-report items, for example, “When I am working, I forget everything else around me,” and “I am immersed in my work.” Items were rated on a 7-point Likert-type scale (1 = Never; 7 = Always).
Perceived Customer Feedback (PCF)
An adequate number of perceived customer feedback measurement items were limited in availability. Therefore, items for this variable were adapted and combined from Frey et al.’s 2013 and Ma and Qu’s 2011 studies. The nine items were rated on a 7-point Likert scale (1 = Strongly Disagree, to 7 = Strongly Agree). Questions included statements such as “I have
received positive customer feedback.” Items were slightly modified to reflect measuring the perceived customer feedback variable more accurately. Ma and Qu’s (2011) measurement reported a Chronbach’s alpha level of .791 whereas Frey, et al.’s (2013) measurement reported .90. Original and modified items are indicated in Table 3.3.
Table 3.3.
Perceived Customer Feedback Measurement Items
Source Original Item Descriptions Modified Item Descriptions Frey, Bayón, &
Totzek, (2013)
I have received positive feedback from this client.
I have received positive feedback from customers.
(continued)
Table 3.3. (continued)
It seems to me that my customers are satisfied with my performance or expertise.
This client has made remarks on my performance or expertise.
Customers have made positive remarks on my performance or expertise.
I have the feeling that my expertise is valued by the client.
I have the feeling that my expertise is valued by customers.
Ma & Qu (2011)
Most guests are polite. Most of my customers are polite.
I feel that my services are
appreciated by our guests. I feel that my services are appreciated by my customers.
I rarely receive complaints from our guests.
I rarely receive complaints from my customers.
I feel our guests are satisfied with the services provided by our hotel.
Customers have told me they are satisfied with the services provided by my organization.
Customers have expressed they are happy with my services.
I feel our guests are happy to stay in our hotel.
Perceived Supervisor Support (PSS)
Eisenberger, Hungtington, Hutchison, and Sowa (1986) developed a 36-item scale known as the “Survey of Perceived Organizational Support” (SPOS). Several articles including Eisenberger et al.’s 2002 study (e.g. Hutchison, 1997a, 1997b; Roades et al., 2001; Kottke &
Sharafinski, 1988) adapted the SPOS to measure perceived supervisor support by replacing the word “organization” with “supervisor” (as cited in Eisenberger et al., 2002). To measure perceived supervisor support (PSS), this study adopted Kottke and Sharafinski’s (1988) questionnaire containing 16 items that reported a coefficient alpha of .98 in the original study. Items were rated on a 5-point Likert scale (1 = Strongly Disagree; 5 = Strongly Agree). An example statement is “My supervisor takes my best interest into account when
he/she makes decisions that affect me.” In addition, the original study contained two items (“If my supervisor could hire someone to replace me at a lower salary he/she would do so”
and “If given the opportunity my supervisor would take advantage of me”) that were reverse coded.
Control Variables
A total of 9 questions were used to collect demographic data; (1) city of employment, (2) frequency of interaction with customers, (3) level of familiarity with customers to conduct work, (4) sector of employment, (5) job title, (6) years with current organization, (7) age, (8) gender, and (9) education completed. Continuous variables (years with current organization and age) were kept in original form during statistical analysis. Categorical and ordinal variables were assigned numeric values in SPSS and dummy coded for use in analyses.
Dichotomous variables (gender) were dummy coded as well with female being 0 and male as 1. For variables with more than two possible responses, some of the most frequently selected answers were chosen for dummy coding, using the same response as 1 and all others 0.
Following suggestions from previous literature (e.g. Blackburn & Bruce, 1989; Clark, 1997;
Hunt & Saul, 1975), two demographic variables which were gender and education completed, also acted as control variables on employee job satisfaction when running SEM.
Validity and Reliability
Validity of the instruments is determined by content validity and construct validity.
Content validity refers to the extent at which a questionnaire reflects the intended domain of the content. That is, the instrument measures what it purports to measure. Content validity is established for all instruments measurements (Schaufeli & Bakker, & 2003; Frey, et al., 2013; Kottke, & Sharafinski, 1988; Ma & Qu, 2011; Weiss et al., 1967) while the instruments were adopted or adapted from the aforementioned studies. The combined items for measuring customer feedback (Frey et al., 2013; Ma & Qu, 2011) underwent expert review while the questions were slightly modified. A pilot test was conducted to evaluate feasibility and
statistical variability of the study (Cleveland & Yeh, 2015). This helped improved the study’s design before the main research project was executed. During the pilot as well as in the main study, exploratory factor analysis (EFA) was used to examine the factor structure and the potential threat of common method variance (CMV). Construct validity was established through the confirmatory factor analysis (CFA) for the main study. The results of EFA and CFA are reported below under their respective headings.
While all items for each survey were self-reported by individual sources in this study, potential threat of CMV needed to be observed. This study used Podsakoff and Organ’s (1986) procedural method recommendations to minimize CMV. The questionnaire items, though still kept within each variable’s domain, were reordered (scale reordering). The first set of questions measured the dependent variable, while the following sets measured work motivation, perceived supervisor support, then perceived customer feedback. EFA results in the pilot test indicated groupings between items in perceived supervisor support and employee job satisfaction, signifying respondents tended to answer items measuring both variables in a similar way. Therefore, in order to reduce CMV in the full-scale study, the questionnaire format was reordered, placing perceived supervisor support at the very end.
Reliability refers to how well an instrument yields stable and consistent results. Studies of which these instruments were adopted from (Schaufeli & Bakker, 2003; Frey et al., 2013;
Kottke, & Sharafinski, 1988; Ma & Qu, 2011; Weiss et al., 1967) have already run reliability analyses. However, a Cronbach’s alpha test was also run for each variable of this study to establish internal consistency of measurements.
Exploratory Factor Analysis
EFA was conducted to observe factor loadings and cross factor loadings of all items within the study. In a simple factor structure all factor loadings were above the recommended threshold of .5 (Fabrigar, Wegener, MacCallum, & Straham, 1999), which was to be expected since all items were previously established measurements.
Using EFA, a Harmon’s single factor test was also conducted to observe common method variance, while all items for each survey were answered by individual sources. The percent of total variances extracted for the largest unrotated factor of the pilot study (43%) suggested possible CMV problems. For the main study, the largest unrotated factor decreased to 37.81% of the variances, which is below the 50% suggested threshold (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Therefore, the test indicated a minimal CMV problem for the full-scale study.
Confirmatory Factor Analysis
After running each variable’s set of items separately through AMOS for CFA, results determined whether modification of items (deleting) was necessary in order to meet model fit cutoffs. When figures did not meet the criteria, modification indices were examined for selection of possible items to delete. After deleting selected items, the modified list of items
After running each variable’s set of items separately through AMOS for CFA, results determined whether modification of items (deleting) was necessary in order to meet model fit cutoffs. When figures did not meet the criteria, modification indices were examined for selection of possible items to delete. After deleting selected items, the modified list of items