The Sustainable Development Goals (SDGs) are a global agenda adopted by the United Nations in 2015 to achieve sustainable and inclusive development. Consisting of 17 integrated goals, the SDGs address a wide range of challenges, from social issues such as poverty alleviation to environmental issues such as climate change mitigation and the adoption of responsible consumption and production patterns.

Green consumer behavior refers to purchasing decisions and consumption patterns that are influenced by environmental concerns and sustainability values. This behavior is driven by environmental awareness and the belief in the effectiveness of individual actions. Within the framework of the Theory of Planned Behavior (TPB), green consumer behavior is explained through attitudes, subjective norms, and perceived behavioral control.

The TPB, proposed by Ajzen (1991), is an extension of the Theory of Reasoned Action (TRA). This theory explains that the intention to perform a particular behavior is influenced by three main factors: attitude toward the behavior, subjective norms, and perceived behavioral control. If an individual has a positive attitude, receives social support, and feels capable of performing the behavior, their intention to engage in the behavior will be stronger.

Purchase intention is the consumer decision-making process that occurs after considering the reasons and needs for purchasing a product or service. This process involves consumer awareness of the product, evaluation of the benefits offered, and comparison with alternative products. Purchase intention reflects the consumer’s willingness to purchase a product based on knowledge, preferences, and the perceived functions of the product.

Figure 4.2 Theoretical Model Image

H1: Environmental knowledge has a positive and significant effect on attitude.

H2: Environmental awareness has a positive and significant effect on attitude.

H3: Attitude has a positive and significant effect on purchase intention.

H4: Subjective norm has a positive and significant effect on purchase intention.

H5:Perceived behavioral control has a positive and significant effect on purchase intention.

H6: Willingness to pay has a positive and significant effect on purchase intention.

Convergent validity is met if the factor loading is ≥ 0.70 (Rahadi, 2023). All indicators meet this threshold, as shown in Table 4.10.

Source: Data processed with SmartPLS, 2025

Table 4.10 shows that all indicators have outer loading values > 0.7. Because they meet validity and reliability requirements, all indicators are declared valid.Discriminant Validity.

Discriminant validity testing ensures that an indicator measures only its own construct, not other constructs. In SmartPLS, this testing is performed using Cross Loadings, the Fornell-Larcker Criterion, and HTMT (Rahadi, 2023).

a) Cross Loading

Discriminant validity can be seen from the Cross Loading. If each indicator has the highest correlation with its construct compared to other constructs, then the indicator is valid as a component of that construct.

Source: Data processed with SmartPLS, 2025

Based on Table 4.12, all indicators have the highest correlation with their own variables compared to other variables. This indicates that all constructs have met discriminant validity.

The Fornell-Larcker discriminant validity test assesses whether the square root of a construct's AVE is greater than the correlation between the constructs. If so, the construct meets discriminant validity (Fornell & Larcker, 1981; Hasheem et al., 2022).

Source: Data processed with SmartPLS, 2025

Based on the table above regarding the results of the Discriminant Validity test using the Fornell-Larcker Criterion method, it can be concluded that each construct has a higher square root mean AVE value than its correlation value with other constructs in the model. This indicates that each construct has good discriminant validity.

The next discriminant validity test uses the Heterotrait-Monotrait Ratio (HTMT), which is considered more sensitive than Cross Loading and Fornell- Larcker (Rahadi, 2023). HTMT compares the correlation between different constructs with the correlation within the same construct. The HTMT value must be <0.9 to indicate good discriminant validity (Henseler et al., 2017).

Source: Data processed with SmartPLS, 2025

Based on Table 4.14, all HTMT values are <0.9, so all constructs are declared valid using discriminant validity using the HTMT method.

Construct reliability testing is conducted by examining the Cronbach's Alpha and Composite Reliability values, which indicate the internal consistency of the indicators in measuring the latent construct (Memon et al., 2017). According to Chin (1998) in Ghozali and Latan (2015), a good value for these two measures is above 0.70.

Source: Data processing results using SmartPLS, 2025

Table 4.15 shows that all constructs have reliability values >0.7, thus being deemed reliable and suitable as measurement tools in this study.

After evaluating the outer model, the next step was to test the inner model to analyze causal relationships among latent variables and ensure that the developed model is robust and accurate. Collinearity was assessed using VIF, with the testing stages carried out as follows:

R-square indicates how well the independent variables explain the dependent variable. A value ≥ 0.75 is considered strong; 0.50 moderate; and 0.25 weak (Effendi et al., 2018). Adjusted R-square provides a more accurate picture as it accounts for the standard error.

Source: Processed data using SmartPLS, 2025

Based on Table 4.16, the R-square value for Attitude (S) is 0.548, with an adjusted R-square of 0.541, indicating that Environmental Knowledge (P) and Environmental Awareness (K) explain 54.8% of the Attitude variable, while the remaining 45.2% is influenced by other factors. This influence is categorized as moderate.

Meanwhile, the R-square value for Purchase Intention (NB) is 0.643, with an adjusted R-square of 0.632, showing that Attitude (S), Subjective Norm (N), Perceived Behavioral Control (KP), and Willingness to Pay (ML) explain 64.3% of the NB variable, with the remaining 35.7% influenced by other factors. This influence is considered strong. Therefore, the structural model in this study is deemed feasible.

The f-square (effect size) measures the impact of exogenous variables on endogenous variables. A value of 0.02 indicates a small effect, 0.15 a medium effect, and 0.35 a large effect; values < 0.02 are considered negligible (Rahadi, 2023). The f² values from the processed data are as follows:

Source: Processed data using SmartPLS, 2025

Table 4.17 shows that the strongest effects are from Environmental Awareness → Attitude (f² = 0.495) and Attitude → Purchase Intention (f² = 0.398). A medium effect is observed for Environmental Knowledge (f² = 0.266), while other variables such as Subjective Norm, Perceived Behavioral Control, and Willingness to Pay exhibit small effects.

The Q² predictive relevance test indicates the model’s predictive capability. A Q² value > 0 indicates the model has predictive relevance, while Q² < 0 indicates otherwise (Kusumawicitra et al., 2022).

Source: Processed data using SmartPLS, 2025

Based on Table 4.18, the Q² values are 0.511 and 0.587. Since the values are greater than 0, the model possesses predictive relevance.

The collinearity statistics test is conducted to determine correlations among independent variables. The model is free from collinearity if the VIF value < 3.3, whereas VIF > 3.3 indicates a violation of the classical assumption (Rahadi, 2023).

Source: Processed data using SmartPLS, 2025

Based on Table 4.19, the variance inflation factor (VIF) values for all indicators are below 3.3, indicating that the data are free from collinearity issues.

The model fit test measures how well the model represents the covariances among indicators. Three measures are used: SRMR, Chi-Square, and NFI. A model is considered fit if SRMR < 0.09 and NFI > 0.5 (Maskur et al., 2015). An SRMR value ≤ 0.08 is still acceptable.

Source: Processed data using SmartPLS, 2025

Table 4.20 shows that the SRMR values are 0.067 (saturated) and 0.081 (estimated), both below the 0.10 threshold, indicating a fairly good model fit. The NFI values are 0.750 and 0.752, both exceeding the 0.50 threshold, thus the model is considered overall fit.

Figure 4.5 Influence Test Model

Table 4.15 Results of Hypothesis Test Analysis

Source: Processed data using SmartPLS, 2023

Based on Table 4.21, all path coefficients have positive values (0.121– 0.511), indicating that all independent variables have a positive effect on the dependent variable. The significance results are shown through the t-statistic and visualized in Figure 4.5.

Based on the statistical calculations in Table 4.21, the following conclusions can be drawn:

Based on the hypothesis testing results, it was found that environmental knowledge has a positive and significant effect on attitude (0.375; p < 0.05), thus H1 is accepted. Consumers who understand the benefits of environmentally friendly products tend to have a positive attitude. This finding is in line with Setyawan et al. (2018) and Kamalanon et al. (2022), which assert that knowledge forms the basis for developing a positive attitude toward sustainability. For Royco’s management, improving consumers’ positive attitudes can be achieved by providing public education, such as including an informative QR code on the packaging that links to sustainability content, and optimizing digital media to communicate Royco’s environmentally friendly values.

Based on the hypothesis testing results, it was found that environmental awareness has a positive and significant effect on attitude (0.511; p < 0.05), thus H2 is accepted. Consumers who are aware of environmental issues have a more positive attitude. This finding aligns with Pangaribuan et al. (2020) and Rivera & Barcellos-Paula (2024), which show that environmental awareness strengthens green identity and attitude.

For Royco’s management, it is recommended to highlight environmental issues in social campaigns and collaborate with green communities to build an environmentally friendly brand image.

Based on the hypothesis testing results, it was found that attitude has a positive and significant effect on purchase intention (0.498; p < 0.05), thus H3 is accepted. A positive attitude encourages consumers to purchase environmentally friendly products. This finding is consistent with Kamalanon et al. (2022) and Kumar et al. (2023), which identify attitude as a key predictor of green product purchase intention.

For Royco’s management, it is recommended to strengthen consumer attitudes through emotional campaigns, customer testimonials, and narratives that purchasing Royco means contributing to environmental conservation.

Based on the hypothesis testing results, it was found that subjective norm has a positive and significant effect on purchase intention (0.211; p < 0.05), thus H4 is accepted. Social support contributes to purchase intention, although not dominantly. This finding is in line with Maichum et al. (2016), which state that social norms play a role in shaping consumer preferences, especially in collectivist cultures.

For Royco’s management, it is suggested to engage influencers and community leaders, as well as develop community-based loyalty programs to strengthen social support.

Based on the hypothesis testing results, it was found that perceived behavioral control has a positive and significant effect on purchase intention (0.194; p < 0.05), thus H5 is accepted. The easier and more affordable the product, the higher the purchase intention. This finding is consistent with Yadav & Pathak (2016) and Paul et al. (2016), which emphasize product accessibility and understanding as factors that encourage purchase intention.

For Royco’s management, it is important to ensure product availability across various distribution channels, simplify information on product benefits, and highlight ease of purchase.

Based on the hypothesis testing results, it was found that willingness to pay more has a positive but not significant effect on purchase intention (0.121; p = 0.068), thus H6 is rejected. The premium price is not yet supported by sufficient perceived benefits. This finding is in line with Kamalanon et al. (2022) and Kumar et al. (2023), which state that willingness to pay is only effective when consumers perceive equivalent benefits. Pinem (2019) and Nassar et al. (2021) also highlight that women tend to be more price-sensitive despite caring for the environment. This is relevant because the majority of respondents (68.5%) were women, who tend to prioritize economic considerations when purchasing environmentally friendly products.

For Royco’s management, it is important to emphasize the tangible benefits of the higher price, for example through education, certification labels, and visualizing contributions to the environment so that the product’s added value is more apparent.