Page 102 - IJPS-11-1

International Journal of
Population Studies Malaysian women in logistics: Stay or switch?

248 samples were collected, the sample size was considered (2022) with a measurement model and a structural model.
sufficient to test the research model of the study. For the measurement model, the convergent validity and
discriminant validity needed to be established before
2.2. Research instruments developing the structural model. According to Ngah

To ensure item validity, all items used to measure the et al. (2022), convergent validity is a test that determines
constructs in the study were adopted from established whether the items used can measure the same construct.
sources in the field. Items for extrinsic motivation and The loading and composite reliability must be ≥0.7, while
intrinsic motivation were adopted from Kuvaas et al., average variance extracted must be ≥0.5 to confirm the
(2017), job satisfaction and family-supportive supervisor convergent validity (Hair et al., 2017). Table A3 of the
behavior from Maloni et al. (2019b), work-life balance from Appendix shows the results of convergent validity.
Fontinha et al. (2018), and items for intention to stay from The heterotrait – monotrait ratio of correlations
Presbitero & Teng-Calleja (2020). Details are provided in (HTMT) was used to establish the discriminant validity,
Table A1 of the Appendix. Since the authors gathered the as recommended by Henseler et al. (2015). As shown in
data from a single source, the risk of common method Table A4 of the Appendix, all HTMT values were lower
variance could arise (Ngah et al., 2022; Tuan Mansor than 0.85, indicating that the discriminant validity was
et al., 2022). As suggested by Podsakoff et al. (2012) and established (Franke & Sarstedt, 2019).
Syafiq et al. (2023), the authors employed different anchor
scales. The exogenous variables were measured using The authors proceeded to the structural model to
5-point Likert scales, whereas the endogenous variable was test the hypotheses after the measurement model was
measured with a 7-point Likert scale. established. As shown in Figure A1 of the Appendix, a
bootstrapping procedure (Hair et al., 2017) with a 5,000
2.3. Statistical analysis resampling technique was used to test the hypotheses
The authors employed the Statistical Package for the Social developed. A hypothesis is considered supported if the
Science (SPSS) to illustrate the respondent’s profile and beta value is aligned with the hypothesis, with a t-value
structural equation modeling (SEM) with smart partial of ≥1.645, a p-value of ≤0.05, and no null values for the
least squares (PLS) (Ringle et al., 2022) for hypotheses confidence interval between lower and upper levels (Ngah
testing. Since the nature of the study was solely for et al., 2021).
predictive purposes, Smart PLS was deemed appropriate 3. Results
for the study (Hair et al., 2019; Ngah et al., 2021).
Furthermore, it has been evidenced that using a single Based on the data obtained from the 248 usable responses,
source of data may result in common method variance most respondents are between the ages of 26 and 33 years
(CMV) (Tuan Mansor et al., 2022), which may have an old (39.1%), and 71.4% of respondents are single women.
impact on the quality of the findings (Halimi et al., 2021). Most respondents have a degree (61.3%). The majority of
To address this problem, both procedural and statistical respondents in the survey are Malays (60.6%). In terms
methods of analysis were applied. A different anchor scale of income levels, the highest proportion of income level
to measure constructs for the procedural method (Ngah is between MYR 1,000 and MYR 3,000, (50%). Exactly
et al., 2021; Podsakoff et al., 2012) was used. As such, a 46.8% of the respondents have between 1 and 3 years of
7-point Likert scale was employed to determine the experience. Table 1 shows the characteristics of the 248
intention to commit. Meanwhile, 5-point Likert scales respondents in terms of their demographic background.
were used to determine the other constructs. For the The authors developed four direct hypotheses and
statistical method, the authors utilized full collinearity four moderating hypotheses. The results of the direct
analysis (Albtoosh & Ngah, 2022; Kock, 2015) to remedy effect in Table 2 show that intrinsic motivation (β = 0.209,
the potential issue of CMV. A VIF value less than or equal p < 0.01), extrinsic motivation (β = 0.158, p < 0.05), JS
to 3.3 indicates that there is no bias in a single source of (β = 0.176, p < 0.05), and family-supportive supervisory
data. The analysis revealed VIF values lower than 3.3, behavior (β = 0.220, p < 0.05) have a positive relationship
indicating that the CMV was not a serious problem in the with the intention to stay in the logistic industry, thus
present study. Table A2 of the Appendix depicts a summary supporting H1 – H4 of the study. This indicates that these
of full collinearity testing for each construct. four variables are crucial factors to increase the intention
to commit in the logistics industry among women in
2.3.1. Validity tests Malaysia. For the moderating effect, the analysis revealed
For the analysis, the authors followed the two-step that two hypotheses were unsupported, and another two
approach proposed by Hair et al. (2017) and Ngah et al. hypotheses were supported.

Volume 11 Issue 1 (2025) 96 https://doi.org/10.36922/ijps.1700