Results

Reliability and validity

An exploratory factor analysis was performed using SPSS 25, and the results are summarized in Table 3. The indicator validity was checked with all factor loadings exceeding a recommended threshold of .70. Internal consistency was confirmed with all constructs' Cronbach's alpha values, and composite reliability values exceeding .70. Convergent validity was established, as the average variances extracted (AVE) were all greater than the acceptable threshold of .5. Finally, discriminant validity was confirmed by comparing the square roots of the AVE values with the corresponding estimates of the correlation values (Table 4). Overall, the measurement items fulfilled the reliability and validity requirements for further analysis.

Table 4 Correlation matrix

Mean SD 1 2 3 4 5 6 7 8 9
Co-creation 5.168 1.016 .811                
Cost-effectiveness 5.676 .940 .556 .920              
Website attractiveness 5.462 .993 .677 .582 .847            
Sustainability 5.102 1.038 .635 .434 .557 .850          
Brand uniqueness 5.378 1.191 .609 .543 .650 .606 .937        
Social media engagement 4.873 1.557 .542 .474 .678 .475 .572 .907      
Innovativeness 5.582 .936 .634 .546 .717 .620 .765 .632 .824    
Attitude 5.554 1.094 .587 .656 .646 .479 .513 .595 .593 .871  
Repurchase intention 5.708 .987 .577 .640 .636 .522 .614 .517 .594 .779 .917
  1. The lower triangle of the matrix represents the correlation coefficients between constructs
  2. The diagonal values (italics values) represent the square root of the average variance extracted of each construct

Measurement model

The data was analyzed through the partial least squares path modeling technique (PLS-SEM), using SmartPLS 2.0 software. As a component-based modelling approach, PLS is often preferred to covariance-based approaches such as structural equation modelling (CB-SEM), and multiple regression when estimating a complex path model with. Our sample size of 210 satisfied the criterion for PLS-SEM with 10 times the largest number of structural paths directed at a particular construct in the structural model, as the sample size threshold for our model would have been 90. Moreover, the use of PLS modeling is recommended when the research model is exploratory in nature, rather than confirmatory. A nonparametric bootstrapping procedure was conducted to test the significance of path coefficients.

The results of the analysis are summarized in Fig. 1 and Table 5. The analysis reveals that the following variables significantly influenced the consumers' attitudes toward DTC brands: co-creation (β = .115, p < .05), cost-effectiveness (β = .480, p < .001), website attractiveness (β = .303, p < .001), brand uniqueness (β = .138, p < .01), social media engagement (β = .300, p < .001), and innovativeness (β = .139, p < .01). Hence, H1a, H2a, H3a, H5a, H6a, and H7a were supported. Sustainability was the only variable that did not have a significant effect on attitude, rejecting H4a. On the other hand, the determinants that significantly influenced consumers' re-purchase intentions include brand uniqueness (β = .331, p < .001), social media engagement (β = .157, p < .01), and innovativeness (β = .115, p < .01), supporting H5b, H6b, and H7b. The variable, co-creation (H1b), cost-effectiveness (H2b), website attractiveness (H3b), and sustainability (H4b), did not significantly affect re-purchase intentions. Additionally, the indirect effects of the independent variables on re-purchase intentions through attitudes were analyzed (see Table 6). While most findings did not differ significantly from the results of the direct effects, it was discovered that the cost-effectiveness variable had a significant indirect influence on re-purchase intentions through attitude (β = .335, p < .001). Finally, consumers' attitudes toward DTC brands had a positive, and significant influence on their re-purchase intentions (β = .700, p < .001), which suggests a strong correlation between attitude and behavioral intentions (H8 supported).

Fig. 1 PLS results of the conceptual model

Fig. 1 PLS results of the conceptual model


Table 5 Results of the Hypothesis Testing

Hypothesis Beta Support
H1a. Co-creation → Attitude toward DTC brand .115* Yes
H1b. Co-creation → Re-purchase intention − .025 No
H2a. Cost-effectiveness → Attitude toward DTC brand .480*** Yes
H2b. Cost-effectiveness → Re-purchase intention .114 No
H3a. Website attractiveness → Attitude toward DTC brand .303*** Yes
H3b. Website attractiveness → Re-purchase intention .082 No
H4a. Sustainability → Attitude toward DTC Brand .072 No
H4b. Sustainability → Re-purchase intention .096 No
H5a. Brand uniqueness → Attitude toward DTC brand .138** Yes
H5b. Brand uniqueness → Re-purchase intention .331*** Yes
H6a. Social media engagement → Attitude toward DTC brand .300*** Yes
H6b. Social media engagement → Re-purchase intention .157** Yes
H7a. Innovativeness → Attitude toward DTC brand .139* Yes
H7b. Innovativeness → Re-purchase intention .115* Yes
H8. Attitude toward DTC brand → Re-purchase intention .700*** Yes
  1. *** p < .001; ** p < .01; * p < .05

Table 6 Results of the indirect effects

Path Beta Support
Co-creation → Attitude → Intention .077 No
Cost-effectiveness → Attitude → Intention .335*** Yes
Website attractiveness → Attitude → Intention .100 No
Sustainability → Attitude → Intention .027 No
Brand uniqueness → Attitude → Intention .231*** Yes
Social media engagement → Attitude → Intention .147** Yes
Innovativeness → Attitude → Intention .118* Yes
  1. *** p < .001; ** p < .01; * p < .05

The percentage of variance explained by the predictors for the endogenous variable of attitude toward DTC brand was 58.0% (R2 = .580). The predictors for re-purchase intentions accounted for 68.7% of the variance (R2 = .687). These R2 values suggest that a high percentage of variance of the endogenous variables was explained, showing support for the conceptualized model.