4.1. Description of Sample Characteristics Distribution
Descriptive statistics of the sample characteristics show that 55.7% of the respondents were male and 44.3% were female, indicating a relatively balanced gender distribution, ensuring the representativeness of gender in the analysis of green consumption behavior. Regarding age distribution, respondents aged 25 to 50 accounted for the highest proportion at 59.8%, with 42.3% specifically in the 25–50 age range. This age group generally has stronger purchasing power and environmental awareness, making them the core demographic for green consumption.
In terms of educational background, 48.2% of the respondents had a high school education or below, followed by 21.4% with a college diploma, 23.8% with a bachelor’s degree, and 6.7% with a master’s or doctoral degree. This result indicates that the sample primarily consists of low to medium-educated individuals, consistent with the general characteristics of consumers, and that differences in educational background may significantly influence green consumption behavior.
Regarding monthly expenditure, 86.2% of respondents reported spending below 10,000 yuan per month, with 38.2% spending less than 2000 yuan, 29.3% spending between 2001 and 4000 yuan, 18.7% spending between 4001 and 6000 yuan, 8.5% spending between 6001 and 8000 yuan, 3.4% spending between 8001 and 10,000 yuan, and only 2.0% spending more than 10,000 yuan. These data reflect that the sample mainly consists of low to medium-income groups, whose consumption behavior may be more price-sensitive. This is particularly important for the subsequent analysis of the acceptance of price premiums for green products.
In terms of occupational distribution, ordinary employees accounted for 19.3%, freelancers for 12.7%, ordinary workers for 15.1%, corporate managers for 8.9%, while students and retirees accounted for 7.5% and 8.4%, respectively. Additionally, individual business owners, homemakers, agricultural workers, government officials, and professionals also accounted for certain proportions. This diverse occupational distribution helps to comprehensively reflect the green consumption behavior of various occupational groups.
Regarding the purchase of green fresh agricultural products, 80.4% of respondents indicated that they had purchased green fresh agricultural products, while only 19.6% had not. This indicates that green products have gained a certain level of popularity and recognition in the market, providing a rich context for consumers’ “identification” behavior during the purchasing process.
In terms of purchase frequency, high-frequency buyers who purchase three or more times per week accounted for 36.5%, with 24.0% buying three times per week, 19.1% twice per week, and 12.5% daily. High-frequency purchasing behavior not only reflects the high demand for green fresh agricultural products but also provides a foundation for studying consumers’ “equilibrium” and “interaction” at different purchase frequencies.
For example, scholars have found that high-frequency buyers, due to prolonged exposure to green products, have stronger recognition of environmental labels and certifications, form a more stable sense of balance between price and quality, and are more likely to engage in interactive promotion through social networks or friend recommendations.
Regarding the awareness and purchase of Zespri kiwifruit, 36.3% of respondents indicated that they had purchased Zespri kiwifruit, 35.9% said they were aware of it but had not purchased it, and only 8.2% said they were unfamiliar with the brand. This indicates that Zespri kiwifruit has a high level of brand recognition in the green fresh agricultural product market [
55].
Due to Zespri’s long-term investment in green certification, sustainable development, and product quality, its products can effectively stimulate consumers’ “identification” behavior regarding green attributes. Additionally, Zespri has enhanced the “interaction” between consumers and green products through brand promotion and social responsibility marketing, further promoting the dissemination of green consumption concepts.
These sample characteristic data provide a foundation for subsequent empirical analysis and strong support for exploring the impact of different demographic characteristics on green purchase intention. The broad coverage of the sample allows this study to comprehensively reflect the characteristics of consumers’ green consumption behavior in the market, which helps to further inform market positioning, pricing strategies, and promotional approaches for green products (see
Table 1).
4.2. Reliability and Validity Analysis
In this study, the analysis was conducted using SPSS Statistics 25.0 and AMOS 23.0. Multiple methods were employed to test the reliability and validity of the scales to ensure the data quality of the measurement results and the effectiveness of subsequent analysis. First, Cronbach’s Alpha coefficient was used to test the internal consistency of each dimension. The value of Cronbach’s Alpha ranges from 0 to 1, with higher values indicating better internal consistency of the scale. Generally, a reliability coefficient below 0.6 indicates the need to redesign the questionnaire or recollect data; a coefficient between 0.6 and 0.7 is considered acceptable, 0.7 to 0.8 indicates high reliability, 0.8 to 0.9 denotes good reliability, and above 0.9 signifies excellent reliability. In this analysis, Cronbach’s Alpha coefficients for all core variables exceeded 0.9, indicating high internal consistency and excellent reliability of the scales.
Meanwhile, to further verify the construct validity of the scales, this study used the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity to assess data suitability and factor structure. A KMO value greater than 0.8 indicates that the sample data are suitable for factor analysis, and Bartlett’s test reaching a significant level (
p < 0.001) suggests good data aggregation properties (see
Table 2).
Under the premise of good reliability statistics and satisfactory Bartlett’s test results, further model fit testing was conducted. According to the model fit indices in the table, the chi-square degrees of freedom ratio (CMIN/DF) is 2.077, which falls within the acceptable range of 1 to 3. The root mean square error of approximation (RMSEA) is 0.051, which is within the acceptable threshold of less than 0.08.
Additionally, the fit indices NFI, TLI, CFI, and GFI all exceed 0.9, indicating an excellent level of model fit. Therefore, based on the overall analysis, the CFA model demonstrates good model fit (see
Table 3 and
Figure 2).
The CFA results indicate that this study employs structural equation modeling (SEM) to validate the measurement model, with the results presented in
Table 4. The standardized factor loadings of the measurement variables are all above 0.70, with only a few indicators slightly below 0.70 but still within the acceptable range.
Most of the variables have factor loadings ranging from 0.70 to 0.88, indicating that each measurement item effectively reflects its respective construct. Among them, the factor loading of GPI1 in green purchase intention (GPI) is the highest (0.879), indicating that this variable contributes the most to the construct.
Overall, the measurement model in this study demonstrates high convergent validity and measurement stability, making it suitable for subsequent structural model analysis.
Additionally, based on Bollen’s (1989) theoretical criteria, the composite reliability (CR) and average variance extracted (AVE) of each latent variable were calculated, and the results showed that all indicators met or exceeded the recommended thresholds, indicating good convergent and discriminant validity of the scales. Therefore, the scales used in this study meet the requirements for reliability and validity in academic research, providing a reliable data foundation for model analysis [
56].
The formulas for calculating AVE and CR are as follows:
Table 4 presents the discriminant validity test results for each dimension in this study. According to the Fornell–Larcker criterion, the square root of the AVE value for each variable is greater than its correlation with other variables, indicating good discriminant validity among the constructs in this study [
57].
Specifically, green purchase intention (GPI) shows a relatively high correlation with interaction (INTRCT) and balance (EQL), with values of 0.438 and 0.412, respectively, suggesting that these two factors play a significant role in predicting green purchase intention. In contrast, environmental regulation (ER) and identity recognition (ID) exhibit lower correlations with green purchase intention, at 0.397 and 0.363, respectively, indicating that these factors have a relatively weaker influence on green purchase intention.
Overall, the measurement model in this study demonstrates good discriminant validity, providing a solid foundation for subsequent structural equation modeling (SEM) analysis (see
Table 5).
The table below presents the descriptive statistics and normality test results for each factor in this study. According to the descriptive statistics, the mean values of all variables range between 3 and 4 (based on a 1–5 positively scored scale), indicating that respondents generally have an above-average level of awareness and behavior regarding environmental regulation and green consumer orientation. In addition, the standard deviation values indicate a moderate degree of data dispersion, reflecting the variability of green consumption behavior among different respondents.
To test whether the data distribution meets normality, skewness and kurtosis were used as test indicators. According to Kline’s criteria [
57], data can be considered approximately normally distributed if the absolute value of skewness is less than 3 and the absolute value of kurtosis is less than 8. Statistical analysis shows that the skewness and kurtosis of all measurement items meet the above criteria, indicating that the data in this study essentially satisfy the requirements for approximate normal distribution, which provides the statistical prerequisite for subsequent structural equation modeling (SEM) analysis (see
Table 6).
In this analysis, Pearson correlation analysis was conducted to explore the relationships among the variables. The analysis results indicate that significant correlations exist among all variables, with all correlations being significant at the 99% confidence level. Based on the correlation coefficients, it is evident that the correlation coefficients
r among all variables are greater than 0. Therefore, it can be concluded that all variables in this analysis exhibit significant positive correlations (see
Table 7).
4.3. Comparison of Hypothesis Testing Results
In the quantitative analysis phase, this study used Structural Equation Modeling (SEM) to validate the path relationships between hypotheses. SEM, as a statistical method capable of simultaneously handling multiple latent variables and their interrelationships, is suitable for exploratory and confirmatory studies of complex models [
58]. Based on the SEM and fit index evaluation criteria proposed by Hair et al. [
59], this study conducted model estimation using AMOS 23.0 software, with a focus on the significance of path coefficients and overall model fit. Model fit evaluation indices include the chi-square to degrees of freedom ratio (CMIN/DF), root mean square error of approximation (RMSEA), and comparative fit index (CFI). All key fit indices met the recommended standards, indicating that the model achieved good fit. This analysis result provides strong statistical support for hypothesis validation (see
Table 8).
SEM Model Path Relationship Hypothesis Testing Results
Path Relationship Hypothesis Testing Results for the Green Consumption Orientation-SEM Model
Recognition significantly and positively predicts interaction (β = 0.29, p < 0.001), thus supporting Hypothesis H1. Recognition significantly and positively predicts purchase intention (β = 0.104, p > 0.001, p = 0.004 < 0.05), thus supporting Hypothesis H2. Interaction significantly and positively predicts purchase intention (β = 0.143, p > 0.001, p = 0.004 < 0.05), thus supporting Hypothesis H3.
Balance significantly and positively predicts recognition (β = 0.275, p < 0.001), thus supporting Hypothesis H4. Balance significantly and positively predicts interaction (β = 0.278, p < 0.001), thus supporting Hypothesis H5. Balance significantly and positively predicts purchase intention (β = 0.115, p > 0.001, p = 0.004 < 0.05), thus supporting Hypothesis H6.
Environmental regulation significantly and positively predicts balance (β = 0.733,
p < 0.001), thus supporting Hypothesis H7. Environmental regulation significantly and positively predicts recognition (β = 0.266,
p < 0.001), thus supporting Hypothesis H8. Environmental regulation significantly and positively predicts interaction (β = 0.316,
p < 0.001), thus supporting Hypothesis H9. Environmental regulation significantly and positively predicts purchase intention (β = 0.155,
p > 0.001,
p = 0.004 < 0.05), thus supporting Hypothesis H10 (see
Table 9).