Page 210 - Hojnik, Jana. 2017. In Persuit of Eco-innovation. Drivers and Consequences of Eco-innovation at Firm Level. Koper: University of Primorska Press
P. 210
In Pursuit of Eco-innovation
Note: Measurement items: Q17A = Reduction in material costs; Q17B = Reduction in pro-
cess/production costs; Q17C = Reduction in costs of regulatory compliance; Q17D = In-
creased process/production efficiency; Q17E = Increased productivity; Q17F = Increased
knowledge about effective ways of managing operations; Q17G = Improved process innovati-
ons; Q17H = Improved product quality; Q17I = Improved product innovations; Q17J = Better
relationships with stakeholders such as local communities, regulators, and environmental gro-
ups; Q17K = Improved employee morale; Q17L = Overall improved company reputation or
goodwill; Chi-square = 613.583; p = 0.000; Goodness-of-fit indexes: NFI = 0.78; NNFI = 0.75;
CFI = 0.79; SRMR = 0.080; RMSEA = 0.216; Reliability coefficients: Cronbach’s alpha = 0.954;
RHO = 0.954; Internal consistency reliability = 0.963.
In the next step, we tried to improve the goodness-of-fit indexes by
conducting another exploratory factor analysis, followed by a confirm-
atory factor analysis. Reduction of items was done step by step; in each
step, we first eliminated the items that showed lower communalities
and had lower correlations with other items (exploratory factor analy-
210 sis). First, we eliminated items that had extracted communalities lower
than 0.60 and correlations with other items below 0.60. After this step,
we again conducted a confirmatory factor analysis to determine wheth-
er the goodness-of-fit indexes had improved, and then we eliminated the
items that had lower standardized coefficients.
Finally, after conducting several exploratory and confirmatory factor
analyses, we came to the best and most parsimonious solution. We have
reduced the number of items from 12 to four. We report further on all the
values from the exploratory and confirmatory factors analyses.
We conducted an exploratory factor analysis by using the overall sam-
ple. The method of extraction in the exploratory analysis was the Maxi-
mum Likelihood Method, while the selected rotation was Direct Oblim-
in rotation, which assumes that different factors are related.
The appropriateness of factor analysis was determined by examining
the correlation matrix of competitive benefits items. The Bartlett’s test of
sphericity, which statistically tests for the presence of correlations among
the underlying variables, showed that the correlation matrix has signifi-
cant correlations (p < 0.05). Furthermore, the Kaiser-Meyer-Olkin meas-
ure of sampling adequacy was examined and indicated similar results;
specifically, the KMO value was 0.810 (KMO value with 12 items was
0.917), which indicates an excellent sample adequacy.
After consideration of each item’s communality index and its contribu-
tion, we retained all the items (the lowest communality after extraction was
0.749, while in the first version with 12 items the lowest communality was
0.511). As noted above, in order to improve the goodness-of-fit we removed
all the items that showed lower communalities (approximately 0.60).
Note: Measurement items: Q17A = Reduction in material costs; Q17B = Reduction in pro-
cess/production costs; Q17C = Reduction in costs of regulatory compliance; Q17D = In-
creased process/production efficiency; Q17E = Increased productivity; Q17F = Increased
knowledge about effective ways of managing operations; Q17G = Improved process innovati-
ons; Q17H = Improved product quality; Q17I = Improved product innovations; Q17J = Better
relationships with stakeholders such as local communities, regulators, and environmental gro-
ups; Q17K = Improved employee morale; Q17L = Overall improved company reputation or
goodwill; Chi-square = 613.583; p = 0.000; Goodness-of-fit indexes: NFI = 0.78; NNFI = 0.75;
CFI = 0.79; SRMR = 0.080; RMSEA = 0.216; Reliability coefficients: Cronbach’s alpha = 0.954;
RHO = 0.954; Internal consistency reliability = 0.963.
In the next step, we tried to improve the goodness-of-fit indexes by
conducting another exploratory factor analysis, followed by a confirm-
atory factor analysis. Reduction of items was done step by step; in each
step, we first eliminated the items that showed lower communalities
and had lower correlations with other items (exploratory factor analy-
210 sis). First, we eliminated items that had extracted communalities lower
than 0.60 and correlations with other items below 0.60. After this step,
we again conducted a confirmatory factor analysis to determine wheth-
er the goodness-of-fit indexes had improved, and then we eliminated the
items that had lower standardized coefficients.
Finally, after conducting several exploratory and confirmatory factor
analyses, we came to the best and most parsimonious solution. We have
reduced the number of items from 12 to four. We report further on all the
values from the exploratory and confirmatory factors analyses.
We conducted an exploratory factor analysis by using the overall sam-
ple. The method of extraction in the exploratory analysis was the Maxi-
mum Likelihood Method, while the selected rotation was Direct Oblim-
in rotation, which assumes that different factors are related.
The appropriateness of factor analysis was determined by examining
the correlation matrix of competitive benefits items. The Bartlett’s test of
sphericity, which statistically tests for the presence of correlations among
the underlying variables, showed that the correlation matrix has signifi-
cant correlations (p < 0.05). Furthermore, the Kaiser-Meyer-Olkin meas-
ure of sampling adequacy was examined and indicated similar results;
specifically, the KMO value was 0.810 (KMO value with 12 items was
0.917), which indicates an excellent sample adequacy.
After consideration of each item’s communality index and its contribu-
tion, we retained all the items (the lowest communality after extraction was
0.749, while in the first version with 12 items the lowest communality was
0.511). As noted above, in order to improve the goodness-of-fit we removed
all the items that showed lower communalities (approximately 0.60).
