This chapter describes the methodology used in this study. The content is divided into the framework of the study, process of the study, research hypothesis, research method, questionnaire design, pilot test, data collection process, measurement, control variables, and data analysis.
Research Framework
In this research, all variables including coming out, workplace friendship, and job satisfaction are presented as dependent variable. The research framework is demonstrated in Figure 3.1.
Figure 3.1. Research framework
Research Hypotheses
Hypothesis 1: Homosexual who are open about their sexual orientation in the workplace are more likely to have higher job satisfaction.
Hypothesis 2: Employees’ perceptions of strong friendship in the workplace will have influence on their job satisfaction.
Hypothesis 3: There is a relationship between Workplace friendship and coming out.
Coming Out
Job Satisfaction
Workplace Friendship
H1 H3
H2
Research Procedures
In the process of the study, the researcher has come up with lots of ideas that is related to her experience and background and tried to find an area which is not only worth studying but also interesting. In the very beginning, the researcher has collected and reviewed the related literature in order to understand the background of the research, discover whether what should be the possible information that will fit and to make the research possible. Identifying the purpose and problems of the study, develop the questionnaire. It takes time to build the framework of the study, but then it is the most important part of the study, it is the best way for people to know not only the topic but also the content. Last but not the least the final conclusion and suggestion. The process of the study will be figured as Figure 3.2.
Figure 3.2. Research procedure
Review the Literature
Identify Research and the Subject
Set the Purpose
Design the Questionnaire
Analyze the Data
Select Sample and Develop
Analyze Data and the Result
Select Sample and Develop Questionnaire Purpose
Provide Conclusions and Suggestions
Participants
Participants were required to be over the age of eighteen to join to the study and must identify himself as a gay or herself as a lesbian, and also have the willing to participate in this research. After, participants must have smartphone phone with Internet access for filling out the questionnaire. The participants in this study will be the Homosexuality that have already come out to their friends, families, or coworkers who have working experience no matter the work experience is full-time or part-time. The first question was set in order to delete those who were not gay and lesbians, which means if they were not our target group, the questionnaire would be stopped by the very beginning. Only these with identified themselves as Homosexuality can they do this questionnaire. The final question listed in the end of the questionnaire is that the participants were asked whether they are in management level or not.
Sampling
The study will utilize non-probable methods of sampling called purposeful convenience sampling. This method includes looking for a person who fills in the questionnaire voluntarily.
Those volunteers who fit the criteria we provided to participate in this study, to find some potential participants if they knew other people that also meet the criteria and may be interested in joining to this kind of research. The advantage of using convenience method is that the researchers will receive a quick response from the participants while the disadvantage is that it will be a lack of diversity of participants.
Data Collection
Two recruitment methods were used in this process: (a) the social networking called Facebook, and (b) email advertisement to personal contacts. Once potential participants learned of the study, they were directed to contact the researcher directly via email with their name and
preferred contact information. Once the information was received by the participant, they were contacted via email or telephone and asked for answering the questionnaire. The Informed Consent Agreement was not only emailed and mailed to all participants, but also conduct a self-addressed stamped envelope. Participants were informed that the agreement needed to be returned in order to participate in the study. Interviews were primarily arranged via email after determining qualifications. The final data collection was conducted through March to May 2018. A total of 630 questionnaires were responded. With a rate of 81.7%, only 515 of the questionnaires were valid in the current study.
Instrument
In this study, a quantitative approach is adopted in order to measure the hypothesis mentioned above; moreover, this research conducted a survey on the Homosexual group to procure the data by means of utilizing questionnaire instrument. After the data was collected, the research would use statistics to measure the hypothesis.
The questionnaire includes questions related to participants’ demographic information and level of disclosure in the workplace. It also contains in total five scales measuring workplace sexual identity management, discrimination, job satisfaction, emotional exhaustion and depressive symptoms respectively. All scales were originally created in English.
Questionnaire Design
This current study will adopt a questionnaire including questions of participants’
demographic information and level of coming out in the workplace. It will also use seven scales to measure the situation of discrimination in the workplace and the part of job satisfaction will adopt five scales to measure their real condition. All scales were created in English and to consider our main target participants will be Taiwanese, so a Chinese version will be made to
those participants that are not so familiar with English in order to ensure they can well-understood all the questions. Also, expert review and pilot test will also be tested after the questionnaire well- organized.
Demographic information. Regarding participants’ background information, their age, gender, education level, whether they are in managing level, and come out or not were in the end of the questionnaire.
Come out. To measure the level of coming out variable in this study adopted Button (1996) to know the level of current coming out situation in the workplace referring to their homosexual orientation. In the very beginning each participant should answer the level of disclosure in the workplace regarding the 4-point scale, 1 means “completely concealed” which means nobody knows about their homosexuality identity in their workplace; 2 represents “incline to conceal”
which means that only those coworkers with a better relationships know about their homosexuality; 3 represents “incline to disclose” which indicates that most of their coworkers know that they are gay or lesbians; and 4 represents “completely disclosed” which almost everyone knows they are homosexual. However, there would be only two choices for them to select. 1 represents “incline to disclose” while 2 means “incline to conceal”. It will be a much easier way for dividing into two groups and prevent the dilemma for those who are completely conceal themselves.
Workplace friendship. Button (2001) proposed one method named, “Treatment Discrimination toward Sexual Minorities Scale”, and it was adopted to grade the level of discrimination of these participants among the workplace including employers, colleagues, and to know the organization’s policies. There are 9 items in the scale with satisfactory internal consistency (α = .97). The participants need to answer according to a 5-point Likert Scale,
ranging from 1 (Strongly Disagree) to 5 (Strongly Agree), and some of the items that were designed in reversed. The meaning of higher scores implies that there is a less level of workplace friendship in current organizational environment. Sample items include “My co-workers are more likely to be supportive of lesbian and gay people because of the training programs maintained by this organization” and “This organization unfairly discriminates against gays and lesbians in the distribution of benefits. For this study, this scale’s Cronbach’s alpha value will be .90.
Minnesota satisfaction questionnaire (MSQ) short form. To measure job satisfaction variable in this study adopted Weiss et al.’s (1967) Minnesota Satisfaction Questionnaire (MSQ) short form, which included twenty items. Items statements of the MSQ inquired participants to rate their degree of their current job’s satisfaction. The questions for example are, “Being able to do things that don't go against my conscience”, “The competence of my supervisor in making decisions”, and “The feeling of accomplishment I get from the job.”
These items would be scored on a 5-point Likert scale (1 = Very dissatisfied; 5 = Very Satisfied). Cronbach’s alpha value from the previous studies had used this measurement stood for the ranges between .85 and .91 (Field, 2002).
Validity and Reliability Pilot Test
A pilot test was conducted with 30 samples, before actual data collection in an attempt to check whether the questionnaire appears problematic to homosexual people in terms of understanding the items, and this method was used to collect data to examine the validity and reliability of the measurement and research design of this study. The result was presented in Table 3.1 The commonly accepted criterion of Cronbach’s alpha is higher than 0.70 (Tavakol
& Dennick, 2011), and the results showed that the Cronbach’s alpha value of all variables were independent variable was tested through confirmatory factor analysis in AMOS. About the related result, the confirmatory factor analysis of workplace friendship (Table 3.2) and CFA measurement model (Figure 3.3) were reported in the following section.
Table 3.2.
χ²= 655.705; df =27; χ²/df =24.285; RMSEA=0.213; CFI=0.812; TLI=0.750; IFI=0.813 Note. N = 515.
For the independent variable, workplace friendship, all of the 9 items were run through CFA analysis. The results of the model fit are presented in Figure 3.3. However, the model was
not so good. In order to reach the criteria, the researcher decided to remove some items (Friendship1, Friendship 4, Friendship 8, Friendship 9) according to modified indices. After modification, the model fit became better. The results of the model fit are presented in Figure 3.4.
Figure 3.3. Measurement model of workplace friendship
Figure 3.4. Measurement model of workplace friendship (modified)
CFA result of job satisfaction. As for the dependent variable, all 20 items were run through the confirmatory factor analysis in AMOS as well. The confirmatory factor analysis of job satisfaction (Table 3.3) and CFA measurement model (Figure 3.5) were presented below.
Table 3.3.
χ² = 1604.918; df =170; χ²/df =9.441; RMSEA=0.128; CFI=0.766; TLI=0.738; IFI=0.766 Note. N = 515.
For the dependent variable, job satisfaction, all of the 20 items were run through CFA analysis. The results of the model fit are presented in Figure 3.5. However, the model was not so good. In order to reach the criteria, the researcher decided to remove some items (Job1, Job2, Job3, Job4, Job5, Job6, Job8, Job9, Job10, Job11, Job12, Job13 and Job16) according to
modified indices. After modified, the model fit became better. The results of the modified model are showed in Figure 3.6.
Figure 3.5. Measurement model of job satisfaction
Figure 3.6. Measurement model of job satisfaction (modified)
All the model fit summaries are demonstrated in Table 3.5. For workplace friendship from the table, it was showed that most of the indices such as CFI, NFI, IFI, TLI, and GFI were all above the criteria. χ²/df and RMSEA were presented as 4.077 and 0.77. On the other hand, in job satisfaction part, CFI = 0.968, NFI = 0.957, IFI = 0.955, TLI = 0.968, and GFI
= 0.963. Also, χ²/df and RMSEA were presented as 3.781 and 0.74 which met the acceptable threshold of the suggested model fit. Please refer to Table 3.4. for more details.
Table 3.4.
The collected data for this research was analyzed through Statistical Package for Social Sciences (SPSS) PC 22.0 statistical software program, and the methods are indicated in the following sections. During this process, the following methods will be adopted including descriptive statistics, confirmatory factor analysis, Pearson correlation analysis.
Descriptive Statistics
First of all, the study will adopt descriptive statistics for the purpose of knowing more general background of the participants. By using this method can we generate some of the results of this study. Moreover, the means, standard deviation, variance will be showed.
Additionally, we will put distribution pattern and percentages of demographic information which demonstrated in this current study, too. Because of the specific participants being chosen
in present study; also, there’s no arbitrary participants joining this study. The main objective of conducting descriptive statistics is to interpret the data concretely such as the meaning, the standard deviation, demographics, and the distribution of the data (Green & Salkind, 2016).
Factor Analysis
This research will adopt Confirmatory factor analysis (CFA) in AMOS, which was conducted to ensure that the measurement is correct in order to make sure the validity and the Cronbach’s alpha value. In this study, we will use CFA to measure if there were some inappropriate questions in pilot test. (Green & Salkind, 2016). Furthermore, CFA will be used to delete some inappropriate questions which existing in the whole questionnaire. Before testing the hypotheses, the researcher would conduct Confirmatory Factor Analysis (CFA) first by using AMOS 22.0 to verify whether our data were fit into a measurement model.
CFA is a widespread statistical technique which has been used for examining the relations among three latent variables. By adopting CFA, it will not only provide a more rigorous standard approach to our research, but also offer a quantitative evidence to let the researcher to ensure internal and external consistency among all measures (Segars, 1997). There is a more important thing that when have to determine model fit, we should really take some common important indices into consideration, such as chi square divided by degree of freedom (χ²/df), RMSEA (root mean square error of approximation), CFI (comparative fit index), and GFI (goodness of fit index). More details about the criteria of suggested model fit are listed in Table 3.5. Fornell and Larcker (1981) proposed that if we want to get a proper construct validity, CR (composite reliability) should be higher than .80 while AVE (average variance extracted) should be higher than .50.
Table 3.5.
Criteria of Suggested Model Fit
Index Threshold
χ²/df 2-5
RMSEA Root mean square error of approximation < 0.08 Good fit
0.08 ~ 0.10 Mediocre fit
CFI Comparative fit index > 0.90
NFI Normed fit index > 0.90
IFI Incremental fit index > 0.90
TLI Tucker-Lewis index =
Non normed fit index (NNFI)
> 0.90
GFI Goodness of fit index > 0.90
AVE Average variance extracted > 0.50
CR Composite reliability > 0.80
Note. Adapted from “Structural Equation Modeling: Guidelines for determining model fit,” by D. Hooper, J. Coughlan, and M. R. Mullen. 2008, Electronic Journal of Business Research Methods, 6(1), pp. 53-60. Copyright 2008 by Dublin Institute of Technology.
Correlation Analysis
It will adopt the correlation analysis to measure the relationship between coming out, workplace friendship, and job satisfaction. The researcher will use correlation analysis to provide the researcher with the correlation coefficients; therefore, researcher can refer the correlation coefficients to judge if the variable changes, the other variable will change as well;
what is more, correlation coefficients can demonstrate the extent of the significance between the variables and the extent of the influence between the variable, also, the relationship among variables, demographics, and control variable. When the correlation is high, it represents that there is a strong relationship between two variables. The figure of correlation should between 1 and -1. When γ = 1 represents perfect correlation, γ = 0, it is no correlation; while γ = -1 means perfect negative correlation. When γ is lower to 0.4, it is regarded as low
correlation; medium correlation is considered γ between 0.4 and 0.7; high correlation is considered γ over 0.7. The correlation coefficient will demonstrate whether variables are related and considered as the base when explaining whether a variable can make predictions of the outcome variable.
Reliability Analysis
To make sure of the reliability of the questionnaire will fit the standard, the reliability analysis was conducted for the consistent results of the data (Green & Salkind, 2016).
Cronbach’s alpha should at least reach a level of .70. to reach the standard.