The sample of donors is composed of 173 respondents, including 87 males (50.3%) and 86 females (49.7%) who belonged to the 18–24 (17.3%), 25–34 (38.7%), 35–44 (18.5%), 45–54 (19.1%), 55–64 (5.8%) and over 65 (0.6%) age ranges. High school is the most common level of education (60.1%). A total of 57.2% of donors are civil servants, and 24.3% are students.
An overview of the sample characteristics is shown in Table III. The other sample included 87 non-donors, of which 65.5% were females and 34.5% were males.
The majority of the sample (52.9%) is in the 25–34 age range. The other respondents belonged to the following age ranges: 18–24 (23%), 35–44 (13.8%), 45–54 (5.7%), 55–64 (2.3%) and over 65 (2.3%). The majority of the sample had a bachelor’s degree (36.8%); 43.7% were civil servants, 34.5% were students, and 10.3% were unemployed.
Table IV shows the distribution of the respondents across Italian regions. The regions with the major number of respondents are Veneto (25.4%), Piedmont (17.9), Lazio (17.9%) and Puglia (11.6).
Table V shows the recurrent motivations for donating the first time and the favourite locations among donors.
Table VI shows the recurrent motivations for donating the first time and the favourite locations among non-donor.
5.2.2 Donors’ and non-donors’ internal reliability and validity
Regarding the donor dataset, the internal reliability of each factor was calculated using Cronbach’s alpha coefficient  and the construct validity was calculated using Convergent Variance Extracted (AVE) and Composite Reliability (CR). Table VII shows the construct validity and reliability values.
Table VII. Construct reliability and validity (Model A, donors)
All the data meet the criteria for acceptable reliability and validity: 0.7 for Cronbach's alpha [52, 53, 54], 0.5 for AVE and 0.7 for CR .
Additionally, for the non-donor dataset, the reliability and validity were calculated using the same measures. The findings are shown in Table VIII.
Table VIII. Construct reliability and validity (Model B, non-donors)
As shown in Table VIII, the data meet the criteria for acceptable reliability and validity [52–54].
5.2.3 Structural Equation Models: multi-group analysis
The conceptual model was tested with SEM using Mplus 7 software.
The adopted procedure is as follows. First, we separately developed models for Group A, i.e., the donors (N = 173), and Group B, i.e., the non-donors (N = 87). Then, we used multigroup analysis to simultaneously identify the main differences between the two independent samples. A graphical representation of the models and the goodness-of-fit indexes are proposed. The robust estimator MLMV was used for continuous variables to correct covariance. Table IX shows the results of the goodness-of-fit parameters. Then, a graphical representation of the measurement models is proposed for both groups. The models are shown in Fig. 4 (Group A, donors) and Fig. 5 (Group B, non-donors).
The results of the SEM goodness-of-fit parameters are presented below.
Root mean square error of approximation (RMSEA = 0.073; 90% C.I. = 0.066;0.080): acceptable according to Browne and Cudeck ;
Critical fit index (CFI = 0.915): acceptable according to Bentler ;
Tucker-Lewis index (TLI = 0.901): acceptable according to Tanaka ;
Standardized root mean square residual (SRMR = 0.062): acceptable according to Hu and Bentler .
Table IX. Goodness-of-fit index model
The analysis confirms that the χ2 (chi-squared) value is significant with its linked probability value. The χ2 test was statistically significant, which indicates an unsuitable fit, even if, according to several authors, it needs to be compared with other indexes before rejection [58, 59, 60, 61].
The other indicators of goodness of fit can be considered adequate since all the values fall within the thresholds suggested by the literature.
Here, the graphical representation of the model is shown for both donors (Fig. 4) and non-donors (Fig. 5).
The graphs represent only the significant relations between factors (p < 0.05).
Figure 4. The conceptual model validated through SEM, Group A, donors
The observed model of Group A shows that there is covariance between Subjective Norm and Perceived Behavioural Control, as in the model previously tested by Ajzen .
Figure 5. The conceptual model validated through SEM, Group B, non-donors
The observed model of Group B shows that there is covariance between Subjective Norm and Perceived Behavioural Control, as in the model previously tested by Ajzen , as well as between Perceived Behavioural Control and Communication.
The results of the two models are summarized below in Tables X and XII. It is possible to notice that the indicators have significant loadings on their assigned constructs. The residual variances are reported in Appendix II.
Table X. Factor loadings statistics, Group A, Donors
Table XI. Factor loadings statistics, Group B, non-donors
The main results and the status of the research hypotheses are summarized in Table XII (Group A, donors) and Table XIII (Group B, non-donors).
Table XII. Status of research hypotheses, Group A, Donors
Regarding the donors, all the proposed hypotheses are supported (p-value < 0.005), except H2, H6 and H7. In particular, Subjective Norm (p = 0.705) does not affect Intention or Inhibitors (p = 0.517). In addition, Communication does not influence Inhibitors (0.271). Attitude (H1) and Perceived Behavioural Control (H3) directly influence Intention. Communication affects Attitude (H2) and Propensity to Generate WOM (H6), which is also positively affected by Intention (H8) and Service Quality (H9).
Table XIII. Status of research hypotheses, Group B, Non-donors
Regarding the non-donors, all the proposed hypotheses are supported (p-value < 0.005), except H1 and H7. In particular, the construct Inhibitors (p = 0.493) is not significant for non-donor respondents. Concerning the construct Attitude, its p-value can be considered marginally significant (p = 0.054), and for the principle of conservation, we decided to accept H1.
Thus, Attitude (H1), Subjective norm (H2) and Perceived Behavioural Control (H3) directly influence Intention. Communication is a predictor of both Attitude (H2) and Inhibitors (H6). Propensity to Generate WOM is predicted by Intention (H8), Service Quality (H9) and Communication (H5).
The observed model in Group A (donors) shows that Attitude (0.441) and Perceived Behavioural Control (0.553) directly and positively influence Intention (H1, H3). Communication has a strong impact on Attitude (1.005) and on Propensity to Generate WOM (0.494) (H4, H5). Propensity to Generate WOM is predicted by Intention (0.216), Service Quality (0.268) and Communication (0.494) (H8, H9, H5). However, Subjective Norm (-0.031) and Inhibitors (-0.025) do not significantly affect Intention to donate (H2, H7), and Communication (-0.066) does not significantly affect Inhibitors (H6) (p > 0.05).
Concerning the observed model in Group B (non-donors), the results reveal that Subjective Norms (0.346), Perceived Behavioural Control (0.410) and Attitude (0.052) affect Intention to donate (H2, H3, H1).
Communication positively influences Attitude (1.000) and Inhibitors (0.183) (H4, H6). Communication (0.505), Intention (0.174) and Service Quality (0.209) affect Propensity to Generate WOM. However, Inhibitors (-0.039) does not affect Intention to donate (H7).