Common method bias can occur due to several sources such as self-reports of the respondents and the use of a single questionnaire for all items to measure constructs of the study (Podsakoff, MacKenzie, Lee & Podsakoff, 2003). It is important to consider this issue because it may affect the research findings. According to Podsakoff et al.
(2003), a traditional technique widely used by researchers to address the common method bias is Harman’s single-factor test. However, this technique has several limitations, such as lack of statistical control of method effects and insensitivity.
Therefore, a multi-factor-method test is suggested to overcome the problem. Both techniques were conducted for the common method bias test in this study.
150 Harman’s single-factor test
Researchers conduct Harman’s single-factor test by loading all of the observable indicators (manifest variables) into an EFA and examining the unrotated factor solution.
A significant level of common method variance presents when (a) a single factor emerges from the EFA or (b) one general factor accounts for the majority of the covariance among the measures (Podsakoff et al., 2003, p. 889).
Table 5.5 indicates the results of the EFA conducted in this study. Twenty-seven indicators (manifest variables) used to measure four latent variables (TLS, RS, MAS and MP) were loaded in to an EFA. The outputs highlight that principal components without rotation determined four distinct components. Thus, the common method bias in the research data is not substantial because more than one factor emerged from the EFA.
151 Table 5.5. Harman’s Single-factor Test
Component Matrix (Non-rotation)
Pattern Matrix (Direct Oblimin rotation)
Component Component
1 2 3 4 1 2 3 4
CH1 0.57 -0.35 -0.05 0.03 0.63 -0.03 0.02 -0.10
CH2 0.67 -0.19 -0.09 -0.07 0.50 0.05 0.04 -0.29
CH3 0.59 -0.48 0.09 0.02 0.72 -0.16 0.16 -0.03
CH4 0.52 -0.41 -0.11 0.12 0.70 0.00 -0.06 0.00
CH5 0.63 -0.45 -0.13 0.19 0.81 0.07 -0.07 0.04
CH6 0.68 -0.45 -0.05 0.16 0.82 0.04 0.03 0.01
CH7 0.72 -0.30 -0.01 0.04 0.66 0.04 0.11 -0.14
CH8 0.63 -0.20 0.16 0.06 0.50 0.05 0.28 -0.06
IS2 0.51 -0.41 -0.20 0.24 0.75 0.12 -0.16 0.10
IS3 0.61 -0.48 -0.10 0.12 0.80 -0.02 -0.04 -0.01
IS4 0.62 -0.46 0.03 0.10 0.77 -0.06 0.10 0.01
RS1 0.59 0.16 0.64 -0.02 0.06 0.08 0.84 -0.03
RS2 0.56 0.14 0.71 0.00 0.05 0.04 0.90 0.03
RS3 0.55 0.18 0.70 -0.05 -0.01 0.03 0.90 -0.03
RS4 0.45 0.16 0.73 -0.04 -0.05 -0.01 0.90 0.03
MAS1 0.48 0.49 -0.15 0.29 -0.06 0.74 0.03 -0.06
MAS2 0.49 0.54 -0.20 0.40 -0.05 0.88 -0.01 0.01
MAS3 0.49 0.39 -0.17 0.41 0.08 0.77 -0.01 0.07
MAS4 0.61 0.44 -0.17 0.29 0.05 0.75 0.03 -0.11
MAS5 0.53 0.41 -0.09 0.42 0.07 0.78 0.09 0.09
MAS6 0.60 0.46 -0.17 0.18 -0.02 0.68 0.04 -0.23
MP1 0.58 0.07 -0.28 -0.47 0.09 -0.05 -0.11 -0.81
MP2 0.68 0.26 -0.22 -0.26 0.05 0.25 -0.01 -0.65
MP3 0.70 0.20 -0.18 -0.41 0.04 0.09 0.03 -0.79
MP4 0.67 0.32 -0.18 -0.34 -0.06 0.21 0.04 -0.74
MP5 0.65 0.20 -0.15 -0.48 -0.01 0.00 0.06 -0.83
MP6 0.57 0.09 -0.14 -0.54 0.01 -0.13 0.04 -0.82
Note. CH = Charisma; IS = Intellectual stimulation; MAS = Management accounting system; MP = Managerial performance; RS = Reward system; TLS = Transformational leadership style.
Extraction Method: Principal Component Analysis
Four components extracted in both non-rotation and rotation
152 Multiple-factor-method test
This approach is the strongest of the approaches depicted by Podsakoff et al. (2003) as it models “method bias at the measurement level”, controls for “measurement error”, and incorporates “multiple sources of method bias” (Podsakoff et al., 2003, p. 897). To apply this approach by using PLS, Liang, Saraf, Hu, and Xue (2007) recommend that a latent method factor should be included in the structural model. This factor is connected to all single-indicator constructs, which are converted from manifest variables of the study. Each latent variable is also linked to the single-indicator constructs converted from its manifest variable. The two incoming path coefficients of each single-indicator construct are similar to the observed indicator loadings on its estimated latent variable and on the latent method factor. These path coefficients can be used to assess if the common method bias presents (Chin, Thatcher & Wright, 2012; Liang et al., 2007).
To assess common method bias, the significances of indicator loadings of the latent method factor need to be examined. Moreover, the variances of each observed indicator explained by its latent variable and by the latent method factor should be compared (Williams, Edwards & Vandenberg, 2003). Squared values of indicator loadings indicate the percentage of indicator variance caused by the construct, the latent variable or the method factor. Evidence of common method bias presents when: (1) the latent method factor loadings are significant, and (2) the indicator variance caused by the latent method factor is greater than that caused by the latent variable (Liang et al., 2007;
Williams et al., 2003).
Figure 5.2 presents the PSL model used to assess the common method bias of the study.
One construct of Common Method and 27 single-indicator constructs converted from 27
153
indicators were added to the structural model in Figure 5.1. Twenty-seven connections from the Common Method to these single-indicator constructs were established. The connections from each latent variable to its single-indicator constructs were also created. The PLS-SEM algorithm and bootstrapping were run with SmartPLS to evaluate the path coefficients, indicator loadings and their significant levels. Table 5.6 exhibits these values. The results indicate that the average variance of the indicators is 0.64, well above the average method factor variance of 0.02, giving a ratio of around 32:1. Moreover, most method factor loadings are not significant. This indicates that common method bias is not a concern.
154 Table 5.6. Common Method Bias Analysis
Construct Indicator
Latent Variables Method Factor
Factor
loadings Significant
Variance explained
Factor
loadings Significant
Variance explained
L L-squared M M-squared
TLS CH1 0.69 *** 0.48 -0.01 0.00
CH2 0.35 ** 0.12 0.38 ** 0.14
CH3 0.83 *** 0.68 -0.10 0.01
CH4 0.81 *** 0.65 -0.16 0.02
CH5 0.92 *** 0.85 -0.16 0.02
CH6 0.87 *** 0.75 -0.05 0.00
CH7 0.56 *** 0.32 0.25 ** 0.06
CH8 0.38 *** 0.14 0.31 ** 0.10
IS2 0.86 *** 0.74 -0.21 * 0.05
IS3 0.89 *** 0.80 -0.14 * 0.02
IS4 0.85 *** 0.72 -0.09 0.01
RS RS1 0.84 *** 0.71 0.09 * 0.01
RS2 0.91 *** 0.83 0.01 0.00
RS3 0.92 *** 0.84 0.00 0.00
RS4 0.93 *** 0.86 -0.11 ** 0.01
MAS MAS1 0.80 *** 0.64 -0.07 0.00
MAS2 0.95 *** 0.91 -0.16 ** 0.03
MAS3 0.75 *** 0.57 -0.01 0.00
MAS4 0.78 *** 0.61 0.09 * 0.01
MAS5 0.78 *** 0.61 0.00 0.00
MAS6 0.70 *** 0.49 0.14 * 0.02
MP MP1 0.86 *** 0.73 -0.09 0.01
MP2 0.71 *** 0.50 0.12 0.01
MP3 0.81 *** 0.65 0.06 0.00
MP4 0.83 *** 0.68 0.01 0.00
MP5 0.87 *** 0.75 -0.03 0.00
MP6 0.82 *** 0.68 -0.08 0.01
Average 0.79 0.64 -0.00 0.02
Note. CH = Charisma; IS = Intellectual stimulation; MAS = Management accounting system; MP = Managerial performance; RS = Reward system; TLS = Transformational leadership style.
*p < .05. **p < .01. ***p < .001.
155 Figure 5.2. Common Method Bias Test in the PLS Model
Note. CH = Charisma; IS = Intellectual stimulation; MAS = Management accounting system; MP = Managerial performance; RS = Reward system; TLS = Transformational leadership style.