Page 210 - rise 2017

The survey questionnaire consists of four major parts: first, communication (COMT) had 4 items
adapted from performance based pay management literature (Anuar et al., 2014; Newman et al., 2016;
Singh, 2009). Second, involvement (INVOL) had 3 items adapted from performance based pay
management literature (Brown et al., 2010; Ismail et al., 2014; McShane et al., 2015; Newman et al.,
2016). Third, performance evaluation (PERFEV) had 4 items adapted from performance based pay
management literature (Ismail et al., 2014; Newman et al., 2016). Four, procedural justice (PROJUST)
had 5 items adopted from (Allen & Mayer, 1990; Meyer & Allen, 1997). All these items were measured
using a 7-item scale ranging from “strongly disagree/dissatisfied” (1) to “strongly agree/satisfied” (7).
Demographic variables were used as controlling variables because this study focused on employee
attitudes.
A purposive sampling technique was utilized to collect 113 survey questionnaires from
employees of the studied organizations. This sampling technique was used because the management of
the organization had not given the list of registered employees to the researchers and this situation could
not allow the researchers to apply a random technique in choosing respondents for this study. The
participants gave their consent prior to answering the survey questionnaires, and it was done on a
voluntary basis.
The PLS-SEM was employed to analyse the survey questionnaire data because it could
deliver latent variable scores, avoid small sample size problems, estimate every complex models with
many latent and manifest variables, hassle stringent assumptions about the distribution of variables and
error terms, and handle both reflective and formative measurement models (Hair et al., 2017). Data for
this study were analysed using the following steps: first, the validity and reliability of instrument were
determined using a confirmatory factor analysis. Second, the structural model was assessed by examining
the path coefficients using standardized betas (β) and t statistics (significant level at t > 1.96). The value
2
of R is used as an indicator of the overall predictive strength of the model. The predictive strength of the
model is determined based on the criteria: 0.19 (weak), 0.33 (moderate) and 0.67 (substantial) (Hair et
al., 2017; Henseler et al., 2009).

Results
The majority respondent characteristics were males (87.6%), aged between 25 to 34 years old (48.9%),
MCE/SPM holders (72.6%), clerical and supporting staff (68.1%), gross monthly incomes from RM2500
to RM3999 (49.6%).
In terms of the validity and reliability of instrument, the values of average variance extracted
(AVE) for COMT, INVOL, PERFEV, and PROJUST were from 0.580 to 0.754 and these values higher
than 0.5, indicating that these constructs met the acceptable standard of convergent validity (Fornell &
Larker, 1981). Besides, the values of AVE square root in diagonal for COMT, INVOL, PERFEV and
PROJUST were from 0.762 to 0.868 and these values greater than the squared correlation with other
constructs in off diagonal. This result showed that these constructs met the acceptable standard of
discriminant validity (Hair et al., 2017; Henseler et al., 2009).
Factor loadings for the items that represent COMT, INVOL, PERFEV and PROJUST were from
0.709 to 0.901. These values stronger on their own constructs, and greater than other items in the
different constructs in the model. This result showed that the items which represent the constructs
respectively met the standard of item reliability analysis (Hair et al., 2017). Further, the values of
composite reliability for COMT, INVOL, PERFEV and PROJUST were from 0.847 to 0.902 and these
values greater than 0.8, indicating that the instrument used in this study had high internal consistency
(Hair et al., 2017).
The mean values for COMT, INVOL, PERFEV and PROJUST were from 4.96 to 5.25 showing
that the levels of all constructs ranging from high (4) to highest level (7). Meanwhile, the values of
variance inflation factor for the relationship between the independent variable (i.e., COMT, INVOLV and
PERFEV) and the dependent variable (i.e., PROJUST) were from 1.429 to 1.527 and these value less
than 5.0, signifying that the data were not affected by serious collinearity problem (Hair et al., 2017).
This result further confirms that the instrument used in this study has met the acceptable standards of
validity and reliability analyses.
The results of PLS-SEM displayed that the inclusion of COMT, INVOL and PERFEV in the
analysis had contributed 17 percent in the variance of PROJUST. This result shows that it provides
moderate support for the model. Further, the outcomes of testing the research hypotheses displayed three