Chapter 7: Farmers’ value-based environmental concern towards the environment
2. Dimensionality and reliability analysis
The hypotheses that Schultz’ EMS formed three distinct dimensions constituted of items measuring the same underlying concept were then tested over the whole sample divided in randomized subsamples of 60% and 40%.10
In both the 60% and 40% portions of the whole sample, the hypothesized three- factor model displayed the best fit to the data (Tables A7.1 and A7.2).
The more complex the model became, the stronger the factor loadings got (Tables A7.3 and A7.4) and the greater amount of variance was explained in Schultz items overall (Tables A7.5 and A7.6). In the three-factor model run on the 60% portion (40%), the factor loadings ranged between 0.835 and 0.978 (0.829 and 0.975), and the amount of variance explained in items, between 0.697 and 0.956 (0.688 and 0.951). Additionally, the correlation between altruistic and egoistic concerns was 0.726 (0.732), between altruistic and biospheric, 0.650 (0.707), and between biospheric and egoistic, 0.573 (0.529).
Cronbach’s alpha for the biospheric, egoistic and altruistic environmental concerns was 0.94, 0.96, and 0.94 respectively. The average inter-item correlations for biospheric, egoistic, altruistic concerns were 0.813, 0.853, and 0.789 respectively. Such values indicate that each scale was composed of items of similar meaning, therefore exhibiting good internal consistency.
b. Comparison between farmers with and without native forest
Next, the dimensionality and the internal consistency of Schultz’ scales were separately assessed on the sample of 291 respondents who did not have native forest fragments on their property and on the sample of 515 respondents with native forest. The responses of the two groups were compared.
10In the 60% portion of the sample, 25 farmers did not answer the EMS question, hence the analysis was
performed on 459 cases; while in the 40% portion of the sample, 20 individuals did not answer, resulting in 302 cases.
Chapter 7: Farmers’ value-based environmental concern towards the environment
i. Farmers without native forest
As for the whole sample, the three-factor model described the data best (Table A7.7). In the three-factor model, the factor loadings (Table A7.8) of the EMS items varied between 0.854 and 0.983, and the amount of variance explained (Table A7.9), between 0.729 and 0.965. As on the whole sample, the three environmental concerns of farmers without forest were highly correlated to each other (Egoistic – Altruistic: 0.740; Egoistic – Biospheric: 0.652; Biospheric – Altruistic: 0.695).
Cronbach’s alpha of the biospheric, egoistic and altruistic environmental concerns scales was 0.93, 0.95, and 0.94 respectively, and the corresponding average inter-item correlations were 0.776, 0.830 and 0.800. As for the whole sample, the scales showed good internal reliability.
ii. Farmers with native forest
The results of the factor analysis of Schultz egoistic, altruistic and biospheric environmental concerns on farmers with fragments also mimicked the results on the whole sample. The three-factor solution described the respondents’ environmental motive concerns best (Table A7.10). Again, the three-factor model exhibited the strongest factor loadings (between 0.803 and 0.974, Table A7.11) and explained the greatest amount of variance in the items (between 0.644 and 0.807, Table A7.12). The three environmental concerns were strongly correlated to one another (Egoistic – Altruistic: 0.726; Egoistic – Biospheric: 0.501; Biospheric – Altruistic: 0.656).
Cronbach’s alpha of the biospheric, egoistic and altruistic environmental concern scales was 0.95, 0.96, and 0.93 respectively. The average inter-item correlation for the three scales were 0.836 (biospheric), 0.866 (egoistic) and 0.784 (altruistic). Again, the three value-based environmental concerns displayed acceptable internal consistency.
Chapter 7: Farmers’ value-based environmental concern towards the environment
iii. Comparison between farmers with and without native forest
Since both farmers with and without fragments responded best and in a similar fashion to the three-factor model, a Multiple Indicators Multiple Causes (MIMIC) analysis was performed on the whole sample (Jöreskog and Goldberger 1975; Muthén 1989) in order to compare the strength of concerns between the two groups of farmers and possible differences in response to particular items. The variable Forest Presence indicated whether respondents had native forest on their property, and was used as the grouping variable. Farmers with forest were expected to hold a stronger biospheric environmental concern, while farmers without forest were hypothesized to hold stronger egoistic and altruistic environmental concerns.
As expected, the three-factor MIMIC model produced an acceptable fit for most of the indices (Table A7.13). The direct effects of Forest Presence on the egoistic, altruistic and biospheric latent factors were not significant, indicating that the two groups of farmers did not differ in their strength of concern towards the three types of objects. Similarly, the modification indices indicated that, holding the latent scores equal between groups, the two groups also did not differ in their response to the individual items. As a consequence, the grouping variable explained none of the variance in the latent egoistic, biospheric and altruistic concerns (Table A7.15). Therefore, farmers with and without fragments exhibited on average the same response to Environmental Motives Scale. The EMS items still loaded strongly on their respective factors (Table A7.14).
c. Comparison between farmers with native forest by choice and farmers with native forest by chance
A MIMIC analysis was then run between farmers who had native forest on their property by choice (74 respondents) and farmers who had native forest by chance (433 respondents).11 Farmers who had native forest by choice were hypothesized to hold a stronger biospheric concern towards the natural environment.
Chapter 7: Farmers’ value-based environmental concern towards the environment
Again, the three-factor model resulted in a satisfactory overall goodness-of-fit (Table A7.16). Farmers with forest by choice and farmers with forest by chance did not significantly differ either in their general level of different environmental concerns, or on their response to particular items. Therefore, the grouping variable again contributed little to explaining the variance of the three latent factors (egoistic, biospheric, and altruistic) (Table A7.18). As before, the various EMS items loaded highly on their corresponding latent factor (Table A7.17).