**4. Results**

In a first descriptive analysis (see Table 4) we can see how the perception of sustainable development, depending on the e ffect of the declaration of the National Park near that village, obtains an average rating (3.60 out of 7). A medium-high perception of tourist activity and visitors is recognised (3.84 and 4.56). An average score is also obtained for the perception of legal limitations on public use associated with the traditional activity of these villages (3.92), in line with the low score given to the question about the increase in wealth (3.23). With respect to the social construct, the item referring to the maintenance of traditions and customs was the most valued (4.21). In the quality of life (QL), an average score was reached by declaring no preference for living elsewhere (4.27); furthermore, the

deficient scores on ease of travel, access to ICTs or actions to respect the environment were highlighted (QL1, QL2, and QL4, respectively).


**Table 4.** Evaluation of the measurement model (starting elements).

A test of normality was then done. The results showed that all variables have a normal distribution. Reliability was evaluated by considering a standardized external load greater or slightly less than 0.70 (see Table 4). The elimination of these indicators resulted in an increase in composite reliability or Mean-Variance Extracted (AVE), as suggested by Hair et al. [67].

The model reliability indicators are shown below, once the elements that do not exceed the reliability cut have been eliminated. The AVE values (defined as the grea<sup>t</sup> average of the square of the indicators associated with the constructions), exceed 0.60, thus demonstrating the convergen<sup>t</sup> validity for all cases. The composite reliability of the 4 constructs is also satisfactory as the values ranged from 0.85 to 0.93 (see Table 5).

**Table 5.** Evaluation of the measurement model (final elements).


Discriminant validity assessed using the criteria defined by Fornell and Larcker [68], which compares the square root of the AVE values with the correlation of the latent variable, was also satisfactory. In fact, as shown in Table 6, the square root of the AVE of each construct is greater than its correlation with any other construct.


**Table 6.** Matrix of correlation between latent variables.

To evaluate the structural model, the R-square for each dependent construct was analysed, as well as the meaning of the trajectories, using Bootstrapping [67]. Figure 2 shows the results of the estimation of the trajectory coe fficients describing the relationships between the di fferent perceptions of the respondents. The standard errors were bootstrapped by considering 2,500 sub-samples, created with observations randomly drawn from the original set of data (with replacement).

**Figure 2.** Estimation of the structural equation model. Notes: ED, Economic Development; SD, Social Development; QL, Quality of Life; GS, Global Satisfaction.

According to the results shown in Table 7, the latent endogenous variables of the model have a weak to moderate explanatory power. The model can explain 21.3% of the residents' perceptions of social development, 51.3% of those related to the quality of life and 83.5% of those associated with the sustainability of the village in terms of public use of the National Park (see Figure 2).

The results of the direct structural relations reveal that all the hypothetical relations are statistically significant, except the one referred to in Hypothesis 2. Four hypotheses are significant at a level of 1% (value *p* < 0.01), hypothesis 6 is significant at 5% (value *p* < 0.05). Social development (SD) is influenced by the quality of life (QL) but not by economic development (ED). On the other hand, QL is strongly influenced by ED (0.723). The results also show the positive and significant e ffects of ED and

QL constructs, with very similar importance (0.448 and 0.447, respectively), and to a lesser extent of SD (0.161).


**Table 7.** Tests of hypotheses for direct effects between latent variables.
