This section characterises and analyses the data collected. Validity and reliability tests were performed. The hypotheses were tested through the PLS-SEM model and the existence of mediation effects was verified. Finally, a multigroup analysis was performed in order to increase knowledge about possible generational differences.
4.1. Sample Characterisation
Regarding the sociodemographic characterisation of the sample, most of the respondents were female (N = 544, 68%), a minority were male (N = 251, 31%) and the rest did not specify (N = 10, 1%). Regarding the age groups of the respondents, the majority were between 18 and 24 years (N = 312, 39%), followed by 25–34 years (N = 178, 22%) and 35–44 years (N = 153, 19%). As for academic qualifications, the largest number of respondents had a bachelor’s degree (N = 367, 46%) and secondary education (N = 241, 30%). Regarding net monthly income, the majority answered that they had no income (N = 228, 28%), which can be justified by the fact that young people represent a large portion of the sample and may only be students. However, many participants reported an income between EUR 700–1399 (N = 181, 22%) and up to EUR 699 (N = 122, 15%).
The main country of origin of the respondents was Portugal (N = 492, 61%), since the questionnaire in Portuguese was disseminated on social networks for a longer period of time. However, through the MTurk platform, responses were obtained from several countries, such as the United States of America (N = 121, 15%) and India (N = 59, 7%).
4.2. Reliability and Validation of Measures
To verify the reliability and validity of the constructs, several indicators were used: the individual reliability of the items, the composite reliability (CR), Cronbach’s alpha α (CA) and the AVE criterion (Average Variance Extracted) [
59,
61].
The results obtained by the PLS-SEM, represented in
Table 1, reveal the individual reliability of the construct indicators and all items reached values elevated to 0.5 [
59,
61]. In this way, no items were deleted. Composite reliability was also achieved, since the values are between 0.863 and 0.956, all higher than the recommended 0.7 [
62]. Considering 0.7 the minimum acceptable value for Cronbach’s α, the reliability of the constructs is confirmed. The values are between 0.765 and 0.939. To measure convergent validity, the AVE criterion was used, which must be greater than 0.5 [
63]. Validity was reached, since the AVE of all constructs (with the exception of frugality) exceeded the appropriate value, ranging from 0.585 to 0.846. Although frugality presents a stroke of only 0.500, as CR is higher than 0.6, the convergent validity of the construct is maintained [
63].
To evaluate the discriminant validity, the Fornell and Larcker criterion was used, which confirms its validity if the square root of the AVE is superior to the correlations of the constructs [
63]. As shown in
Table 2, the square root of the AVE for each variable (indicated in bold) is greater than the correlations between the remaining constructs. Thus, according to this criterion there is discriminant validity.
In summary, reliability, convergent validity and discriminant validity have been proven. The model has been validated and it has been confirmed that the scales used are adequate to measure the respective variables.
4.3. Hypothesis Testing
After verifying the validity of the items, hypothesis testing was performed using the bootstrapping procedure with 5000 samples.
Table 3 presents the results obtained for the coefficients, the values of
t and
p.
The results of the coefficient of determination show that the various antecedents explain 31.2% of the variance of attitude, 19.9% of the variance of the subjective norm and 29.9% of the variance of the perception of behavioural control. In turn, the antecedents and constructs of TPB (attitude, subjective norm and perception of behavioural control) explain 63.9% of the variance of purchase intention.
H1 proposes that attitude (H1a), subjective norm (H1b) and PBC (H1c) positively affect the intention to purchase secondhand products. The results validate the positive relationship for the three constructs, and attitude (coef = 0.395, p < 0.001) plays a higher role than the subjective norm (coef = 0.087, p < 0.05) and PBC (coef = 0.270, p < 0.001).
H2, which proposes that price (PR) positively influences attitude (H2a) (coef = 0.439, p < 0.001), PBC (H2b) (coef = 0.502, p < 0.001) and directly intention (H2c) (coef = 0.186, p < 0.001), is supported.
H3 proposed that the need for exclusivity (NE) positively influences the subjective norm (H3a) and intention (H3b). However, NE is not significant in intention only about the subjective norm (coef = 0.437, p < 0.001). Thus, only H3a is empirically validated.
H4 proposes that frugality positively affects the three constructs of TPB and intention. As shown in
Table 3, the effect of frugality is not significant in the subjective norm but is in attitude (coef = 0.150,
p < 0.001), PBC (coef = 0.115,
p < 0.01) and intention (coef = 0.067,
p < 0.05). Therefore, H4a, H4c and H4d are validated and H4b rejected.
H5 proposed that environmental concern (EC) positively influences the different behavioural variables of TPB and intention. However, EC only significantly affects attitude (coef = 0.143, p < 0.05) and intention (coef = 0.115, p < 0.01). In this way, H5a and H5d are confirmed. The results reject H6, which means that the effect of environmental knowledge (EK) is not significant on attitude (H6a), PBC (H6b) and intention (H6c).
To test the existence of mediation effects, indirect effects and adjusted confidence intervals were extracted and represented in
Table 4. In the theoretical model described (
Figure 1), attitude, SN and PBC can generate this mediating effect between the antecedent variables and the intention to purchase secondhand products.
Starting with the mediation effect of the attitude between price and intent, the IC for the indirect effect does not include zero (95% IC [0.128, 0.226]). The same happens in the PBC’s mediation effect between price and intent, the IC for the indirect effect does not include zero (95% IC [0.102, 0.177]). The results show that attitude and PBC mediate the relationship between price and intention and as the direct effect is also significant (coef = 0.186,
p > 0.001), in both cases, it is a complementary mediation [
64].
On the mediation of the subjective norm between the need for exclusivity and intention, the IC for the indirect effect excludes zero (95% IC [0.015, 0.064]). It was found that SN mediates the relationship between NE and intention, but since the direct effect is not significant (coef = −0.005,
p > 0.05), it is only an indirect mediation [
64].
Attitude (95% IC [0.028, 0.095]) and PBC (95% IC [0.011, 0.055]) mediate the relationship between frugality and intention, since IC for the indirect effect exclude zeros. As the direct effect is also significant (coef = 0.067,
p < 0.05), it is a complementary mediation [
64].
However, the same is not true in the mediation of the subjective norm between frugality and intention. As the IC for the indirect effect includes zero (95% IC [0.000, 0.014]), SN does not measure the relationship, but the direct effect is significant (coef = 0.067,
p < 0.05), so it is a direct effect without mediation (direct only non-mediation) [
64].
Regarding the mediation effect of the attitude between environmental concern and intention, the IC for the indirect effect excludes zero (95% IC [0.023, 0.094]), showing that attitude mediates the relationship. As the direct effect is also significant (coef = 0.115,
p < 0.05), it is a complementary mediation [
64].
In the case of IC for the indirect effect of the subjective norm (95% IC [−0.001, 0.013]) and PBC (95% IC [−0.014, 0.034]), these include the zeros and therefore they do not mediate the relationship but the direct effect is significant (coef = 0.115,
p < 0.05). Therefore, it corresponds to a direct-only non-mediation effect [
64].
Finally, in the case of environmental knowledge, attitude and PBC do not mediate the relationship between EK and intention, since both IC for indirect effects include zeros (95% IC [−0.031, 0.036] and (95% IC [−0.047, 0.005]), respectively. In addition, the direct effect is also not significant (coef = 0.022,
p > 0.05). Therefore, there is no effect (no effect non mediation) [
64].
4.4. Multigroup Analysis
To test the results of different groups, a multigroup analysis was conducted, although this analysis did not correspond to the main objective of the study. The data were grouped in order to obtain only two groups. The first group is young people between 18 and 34 years old (young), the second is composed of individuals over 34 years old (old). This analysis was carried out in order to verify possible behavioural differences between individuals of different ages and which factors have the greatest influence on the intention to purchase secondhand products in each age group.
The estimated coefficients of the three groups and the estimated differences between them were recorded in
Table 5.
The results suggest some significant differences between the estimated coefficients of the two groups. The main differences were between attitude and intention (p < 0.05), price and attitude (p < 0.05), price and perception of behavioural control (p < 0.05), frugality and subjective norm (p < 0.05), environmental concern and attitude (p < 0.001).
Attitude has a greater impact on purchase intent for younger consumers (coef = 0.446, p < 0.001) compared to older consumers (coef = 0.285, p < 0.001). The former believe more in the positive impact of secondhand buying than the latter. For older people, price has a greater impact on attitude (coef = 0.548, p < 0.001), while for younger people, its impact is less strong (coef = 0.362, p < 0.001). In the case of the effect of price on PBC, the effect is greater on older individuals (coef = 0.576, p < 0.001) than on young people (coef = 0.446, p < 0.001). Regarding frugality, there were also different effects. Its effect on the subjective norm is significant (coef = 0.168, p < 0.001) in the elderly, but not in the younger.
Finally, the effect of environmental concern on attitude was significant only in younger people (coef = 0.289, p < 0.001), and not in the case of individuals older than 34 years.
The results allow us to infer the impact of different factors on the mediation effect of behavioural variables and generational differences.
Table 6 presents the validation of twelve hypotheses and the rejection of seven hypotheses.