This chapter describes detailly the methodology using in this research including: the research framework, hypotheses, target participants, instrumentation and data analysis. The study using quantitative research process to test the relationship among the variables.
Research Framework
The research framework is presented below to provide the relationships among three main variables in this study. Based on the literature review, the independent variable was protean career orientation with two dimensions: self-directedness and value-driven orientation.
The dependent variable was intention to stay, and perceive organizational learning culture was treated as a moderator.
Figure 3.1. Research framework Protean Career
Orientation - Self-directness
- Value-driven Orientation
Perceived Organizational Learning Culture
Intention to Stay
Control Variables Age
Gender
Organizational Tenure H1 (+)
H2 (+)
22
Hypotheses
According to the purpose of the research, literature review in previous parts and research framework, this study proposed the hypotheses as following:
Hypothesis 1: Protean career orientation is positively associated with intention to stay Hypothesis 2: Protean career orientation and intention to stay is positively moderated by perceived organizational learning culture; this relationship is strengthened when perceived organizational learning culture is high.
Research Procedure
The research procedure in this study was demonstrated in seven stage. Firstly, the researcher found interests in talent retention in organization as it has been a critical problem in Vietnam during the observation for a long time working in Vietnam. After figuring out the topic of interest, the researcher studied more literature review in order to specify the topic. In particular, literature review of factors which lead to talent retention in organization.
Secondly, research framework and hypotheses were developed after the literature review was conducted.
Thirdly, the research methodology was designed to implement this study. In this study, the quantitative study with purposive sampling method was selected to be carried out. After the research method was designed, the next step was to employ a measurement scale for this research. The questionnaires were chosen based on the empirical studies relevant to the topic of this study. Appendix section provides the detail questionnaire used in this study.
Data collection was implemented afterward. The target participants in this study were the millennial workers in Vietnam. Questionnaires were distributed through the online survey and were collected through the convenient sampling. Specific reasons for using this sample method were deliberated in the data collection part.
23
As soon as the data collection was implemented successfully and the researcher collected enough data for the study, data was analyzed to find the results. In data analysis stage, the researcher firstly intented to code the data using Microsoft Excel, and used SPSS version 23.0 to analyze the data. At the same time, AMOS version 23.0 was used to analyze the confirmation factor analysis in this study.
Finally, findings and discussion of the hypothesis results were demonstrated at the end of this study. The researcher also came with the conclusion and suggestion for future research about the relevant topics.
Figure 3.2. Research procedure
Findings and final report Data analysis
Data collection implementation Measurement instrument selection
Research method design
Research topic identification and relevant literature review
24
Sample
The research targeted to millennials and a part of generation Z, who were born from 1981 to 2000 (ages of 19-39) and are in Vietnam now. According to a report conducted by Navigos Group (2017), millennials and gen Z are the major workforce in Vietnam. They were calculated at 27 million people in Vietnamese population, equalling to thirty percent of the total population in Vietnam. It was accounted for 50% of Vietnmese workforce based on the research of Labour and Social Trends in Vietnam 2012-2017 carried out by International Labour Ogranization.
Measurement
The measurement of this study was applied from the empirical researches related to protean career orientation, intention to stay, and organizational learning culture.
Protean Career Orientation
As mentioned in the previous part, protean career orientation played the role as an independent variable in this study. In this study, the researcher applied the measurement of protean career orientation, including 14 items developed by Briscoe and Hall (as cited in Briscoe, Hall, & DeMuth, 2006). The measurement was built with the attendance of Hall, the owner of proetan career concept, so that there was a consistence through the research of protean career orientation.
The measurement scale consisted of two dimensions. There were 8 items to measure self-directness and 6 items to measure value-driven orientation. The measurement was designed in five-point Likert scale in which 1 equal “strongly disagree” and 5 means “strong agree”. The participants responded their opinions following the scales to express their degree of protean career orientation.
For self-directed career management, some sample items were “When development opportunities have not been offered by my company, I have sought them out on my own”. The
25
sample item of value-driven orientation was as “I navigate my own career, based on my personal priorities, as opposed to my employer’s priorities”. The Cronbach’s alpha for the protean career orientation was .82.
Perceived Organizational Learning Culture
In this research, perceived organizational learning culture was the moderator. To measure organizational learning culture, Marsick and Watkins (2003) developed Dimensions of the Learning Organization Questionnaire (DLOQ). The original questionnaire obtained 43 items to assess 7 dimensions of organizational learning culture. However, Yang (2003) tested the validity and reliability of the original DLOQ measurement, and the scholar recommended to use a full version of this measurement only if “practitioners want to use the DLOQ as a diagnostic tool” (p.160); or it should use a short version of this questionnaire containing 21 items to investigate the relationship between organizational learning culture and other variables.
In particular, the short version has 3 items in each dimension deleted from the original version.
This short version has been validated through many previous studies (e.g. Egan et al., 2004;
Park & Rothwell, 2009; Joo, 2010; Joo & Shim, 2010). According to Yang’s (2003) questionnaire version, the Cronbach’s alpha estimated for 21 items reached at .93. Therefore, this study employed the short version to assess perceived organizational learning culture. In this study, the Cronbach’s α coefficient for this scale was as high as .93.
The samples of this measurement in individual, team and organizational levels were
“In my organization, people are rewarded for learning”, “In my organization, teams/groups have the freedom to adapt their goals as needed”, and “My organization makes its lessons learned available to all employees”. Participants was questioned to clarify the degree to which they perceive whether their organization has learning culture or not (1 = almost never; 7 = almost always).
26
Intention to Stay
The measurement of intention to stay was derived from 4-item scale created by Kim, Price, Mueller, and Watson (1996). Respondents were asked to rate on a seven-point scale from 1 as not at all agree to 7 as strongly agree to express their opinion of stay intention. The sample item for this scale was “I would be reluctant to leave this organization”. In addition, the reliability of this scale proven by Cronbach’s alpha score was as .71.
Control Variables
Following the empirical literature on the protean career orientation and intention to stay, some demographic variables such as age, gender, organizational tenure influenced on the intention to stay. Thus, this study controlled these variables to ensure the relationship among the main three variables that would not be impacted by these demographic variables.
Age
The older workers had intention to stay longer than younger workers (Arnold &
Feldman, 1982). Other researches of intention to stay also considered age as their control variables with the beliefs as age was related to stay intention (e.g. Chew & Chan, 2008; Huang, Lin, & Chuang, 2006), and the evidence showed that age had the negative effect on intention to leave (Sager, Futrell, & Varadarajan, 1989). Hence, age was treated as a control variable which was coded into a range as 1 for 19-25 years old, 2 for 26-30 years old, 3 for 31-35 years old, 4 for 36-40 years old.
Gender
Price and Kim (1993) figured out that females manifested a significantly stronger intention to stay than males. Other studies about the intention to leave also argued that men less stick to their job or less stayed in a single organization than women (Stocks, 1997; Kanter, 2008). Therefore, gender was a control variable in this study and was coded as 1 for male and 2 for female
27
Organizational Tenure
The study from Price and Kim (1993) elicited that the longer employees worked in an organization, the much intention to stay with that organization employees had. It implied that organizational tenure affected employees’ tendency to leave or stay in a company. As this reason, the researcher put organizational tenure as a control variable in this study. An open-ended question was designed to collect the data of organizational tenure.
Questionnaire Design
Initially, all measurements used in this research were well-developed to guarantee the validity of the research. The measurement of each variables had been applied in previous studies, and the measurement scale with high Cronbach’s alpha was applied to develop the questionnaire. In addition, Podsakoff, MacKenzie, Lee, and Podsakoff (2003) stated that one of the methods to reduce the biases was to develop a questionnaire with different scales among variables. Based on those ideas, the researcher built the questionnaire with five-point Likert scale for the independent variable and seven-point Likert scale for the dependent variable. For the moderator, seven-point Likert scale was utilized.
Next, as the original scales were all developed in English, translating those questions into Vietnamese was necessary because of the targeted participants. To assure Vietnamese version accurate, and to maintain the right meaning from the original ones, several steps of back translation (Brislin, 1970) was implemented. First, all items in the questionnaire were translated to Vietnamese by an English teacher who are fluent in both English and Vietnamese academically. After English-Vietnamese translation was completed, the researcher asked for two Vietnamese students whose English ability is proficient, and their English level was demonstrated by the IELTS score at 7.0 or above, to translate back the Vietnamese version into English. By that way, it was possible to check whether the translated version keep the exactness of the terms of the original version.
28
Although this research went with the different scales in the questionnaire to reduce the common method variance, it was argued that research purpose also guided the participant to answer and led to the bias (Podsakoff et al., 2003). Thus, in the questionnaire, the researcher did not reveal the purpose of the research.
Finally, the reverse questions were brought into the survey aiming to prevent the random answers from the respondents.
Data Collection
In the period of data collection, 288 samples were collected from the targeted employees working in Vietnam regardless of the industry, and there were 253 valid samples.
The participants reported themselves in the questionnaire because the researcher targeted to the individual level in order to test the association among protean career orientation, intention to stay and perceived organizational learning culture.
Moreover, the purposive sampling method and snow-ball sampling method were employed to collect the data. In the implementation of collecting data, the questionnaire was designed on an online survey platform, and the researcher started to distribute this survey. In the beginning, the researcher identified the potential candidates who were employees working in the office in Vietnam, and sent the survey to those targets through some social media platforms such as Facebook, Linkedin, Zalo, etc. The researcher also sent the targeted respondents an email to ask the assistance to fill the survey. Afterward, the participants forwarded to other people. In this way, the questionnaires were spread out the population.
The process of data collection was operated within 1.5 months from the first questionnaire was disseminated since the mid-February, 2020.
29
Pilot Test
The studier collected a small group of 59 targeted participants to conduct a pilot test for the purpose of validating the reliability and consistency of this research before distributing the official questionnaire publicly. The analysis of pilot database was implemented in the earlier of February, 2020.
In the first step, demographic information was analyzed by the SPSS software. Among the samples, there were 40 females and 19 males, and 79.7 percent of them are from 26 to 30 years old. Bachelor degree holder were 48 (81.4 percent), and the rest hold a master degree (18.6 percent). There were no people possessing a high school degree or getting a doctoral degree. As all of the respondents were working in full time but at a young age, most of them stayed at an executive position (38 people equals to 64.4 percent of the whole). For the organizational tenure, the average working years in the current organization was approximated to 2 years 4 months.
Next, Cronbach’s alpha was investigated to assure the reliability for certain. Nunnally (1978) asserted that reliability was acceptable if the Cronbach’s alpha reached to .70. In this pilot test, Cronbach’s alpha score of protean career orientation was at .71, perceived organizational learning culture got .88 meanwhile the intention to stay reached the Cronbach’s value at .63. Although, the coefficient value of the intention to stay did not reach .70, the scale could be used according to Hair, Black, Babin, Anderson, and Tatham (2006). The scholars expressed that a scale could be accepted near of .60, especially when the factor had only few items. In this research, there were only four items in the scale of intention to stay, so that the result of Cronbach’s alpha was understandable. Following, Pearson correlation analysis was adopted to test the correlation among the variables. Table 3.1 reported detailly the analysis of Pearson correlation and Cronbach’s alpha coefficient values.
30 Table 3.1.
Means, Standard Deviations, and Intercorrelations among study variables (n=59)
Variables Mean SD 1 2 3 4 5 6
1. Gender 1.68 .47
2. Age 2.03 .56 -.09
3. Organizational
Tenure 2.28 2.59 .057 .24
4. Protean Career
Orientation 3.94 .36 -.04 .093 -.03 (.71)
5. Organizational
Learning Culture 5.03 .65 .11 -.05 .02 .08 (.88)
6. Intention to Stay 4.6 .77 -.12 .15 .08 .40** .42** (.63) Note. Alpha internal consistency reliability coefficients appear on the main diagonal.
Gender: Male = 1, Female = 2; Age: 19-25 = 1, 26-30 = 2, 21-35 = 3, 36-40 = 4
*p < .05, **p < .01
Protean career orientation positively correlated with intention to stay (r = .40, p < .01), and organizational learning culture also possitive associated with intention to stay (r = .42, p < .01). Besides, none of the control variables were related to protean career orientation, organizational learning culture and intention to stay. In general, the positive correlation among the variables was signified, which had brought an expectation of supported results in this research.
31
Data Analysis
As soon as the data collection was enough for the purpose of doing a quantitative research, the next step of the research was to analyze the database by the SPSS software version 23.0 and AMOS version 23.0. The data analysis section contained several parts as following:
Common Method Variance (CMV) Analysis
Common method variance affected item reliabilities and validities (MacKenzie &
Podsakoff, 2012). CMV has been known as a “systematic error variance due to characteristics of the specific method being employed which may be common to measures of other traits/constructs” (Campbell & Fiske, 1959, as cited in MacKenzie & Podasakoff, 2012, p.542).
Therefore, Harman’s One-Factor Test was availed in order to detect whether the CMV existed in this study.
According to Podsakoff and colleagues (2003), a study is considered to be affected by CMV when a single factor consists of over 50 percent of the sums of squared loadings. As such, the researcher entered all items of the whole variables to identify the existence of CMV. Table 3.2 displayed the result of Harman’s Test, 26.46 percentage of the first factor was estimated. It manifested that the CMV did not significantly get impact on this research.
32 Table 3.2.
Total Variance Explained (39 items)
Total Variance Explained
Component Initial Eigenvalues Extraction Sums of Squared Loadings Total % of
33 Table 3.2 (Continued)
20 .603 1.54 81.86
21 .584 1.49 83.36
22 .549 1.40 84.76
23 .509 1.30 86.07
24 .476 1.22 87.29
25 .456 1.17 88.46
26 .452 1.15 89.61
27 .427 1.09 90.71
28 .396 1.01 91.73
29 .391 1.00 92.73
30 .380 .97 93.70
31 .341 .87 94.58
32 .321 .82 95.40
33 .308 .79 96.19
34 .298 .76 96.96
35 .273 .69 97.66
36 .265 .68 98.33
37 .239 .61 98.95
38 .210 .54 99.49
39 .199 .51 100.00
34
Convergent Validity
This study examined the construct study with a convergent validity. Convergent validity refers to the degree to which different measures are related to each other (Campbell
& Fiske, 1959). The convergent validity can be measured by two criteria including composite reliability (CR) and average variance extracted (AVE). From the research of Fornell and Larcker (1981), AVE measures the level of variance captured by a construct versus the level due to measurement error. The value of AVE is considered good if it is 0.7 and above;
whereas 0.5 is considered as acceptance. For CR, value is acceptable as high as 0.6 (Fornell &
Larcker, 1981). Therefore, convergent validity is determined if the variables of the study have CR values greater than 0.6.
Table 3.3
CR and AVE values of Protean Career Orientation
Variable No. Error variance Factor Loading T Value CR AVE
35 Table 3.4.
CR and AVE values of Organizational Learning Culture and Intention to Stay
Variable No. Error variance Factor Loading T Value CR AVE
36
The Table 3.3 and Table 3.4 above displayed the CR and AVE values of three main constructs used in this study. From the report, all AVE values stayed in an acceptable range which closed to 0.5, and the values of CR were above 0.7 confirming the convergent validity of this study.
Cronbach’s Alpha
Cronbach’s Alpha coefficient was operated to examine how the stability and consistency of the measurements in this study were indicated. Alpha coefficient value ranges from 0 to 1. The higher the value is, the higher reliability of a measurement scale is (Santos, 1999). Also stated from Nunnally (1978), Cronbach’s alpha was at 0.7 which would prove an acceptable reliability. In this research, the protean career orientation was calculated at .82, referring to good enough for the scale’s reliability. For perceived organizational learning culture, the value was as high as .93, it meant that the scale had a good reliability. Intention to stay’s Cronbach’s Alpha also reached to an acceptable level at .71. The summary of Cronbach’s alpha values of each variable was displayed in the Table 3.5.
Table 3.5.
Cronbach’s Alpha of Measurement (N= 253)
Variables Cronbach’s Alpha
Protean Career Orientation .82
Perceived Organizational Learning Culture .93
Intention to Stay .71
Confirmatory Factor Analysis (CFA)
The implementation of confirmatory factor analysis was to test the validity of the questionnaire. Furthermore, the confirmatory factor analysis supported the researcher to confirm how well the data fit the hypothesized model through groups of participants and times (Brown, 2015). From the view of Newsom (2020), several catogories of fit indexes include
37
absolute fit indexes, relative (or comparative) fit indexes, parsimony fit indexes. The evaluation of a model fit should report one of those fit indexes (Hair, Anderson, Babin, & Black, 2010) including one absolute fit index which was represented by standardized root mean square residual (SRMR) or root mean square rrror of approximation (RMSEA), one of relative fit indexes as incremental fit index (IFI) , normed fit index (NFI), comparative fit index (CFI).
In order to assure a model fit, some standard of fit indexes have to be attainted (Hair et al., 2010). Hu and Bentler (1999) did work cutoff critea for fit indexes. In particular, they recommended to use one of the relative fit indexes (e.g. NFI, IFI) above .90 while comparative fit index (CFI) should be greater than .90, in addition to SRMS and RMSEA close to .06 or around .08. For the square, in spite of some arguments existed among the scholars as Chi-square depends on the amount of samples (West, Taylor, & Wu, 2012), the ratio of X2/df was introduced to indicate a good fit if a value of 5 or less (Jöreskog, 1969, as cited in West, Taylor,
& Wu, 2012).
The particular criteria of fit indexes were summarized in the Table 3.6.
Table 3.6.
Summary of Goodness of Fit Indexes
Index Threshold References
X2/df < 5.00 Jöreskog, 1969; West,
Taylor, & Wu, 2012
SRMR < .08 Hu & Bentler, 1999
RSMEA < .08 Hu & Bentler, 1999
CFI > .90 Hu & Bentler, 1999
IFI > .90 Hu & Bentler, 1999
CFA result of Protean Career Orientation
The confirmation factor analysis for the independent variables was conducted firstly by AMOS software. Protean career orientation with 14 items was divided into a second-order
38
construct presented two dimensions (self-directness and value-driven) to analyze. The measurement was proved to be consistent, refer to Figure 3.3. The value of chi-square was
construct presented two dimensions (self-directness and value-driven) to analyze. The measurement was proved to be consistent, refer to Figure 3.3. The value of chi-square was