**3. Methodology**

#### *3.1. Measurement Instrument*

The questionnaire had three parts, namely, (1): socio-demographic characteristics of consumers (e.g., gender, age, education level, personal income, and place of living); (2) general questions about consumers' sustainable consumption behaviour; (3) factors influencing consumers' sustainable consumption behaviour. All measurement scales for the constructs have been included in prior publications. The SCB scale is a multi-dimensional second-order construct, which incorporates three dimensions, i.e., "Quality of Life (QL)", "Care for the Future Generation (CEW), and "Care for the Environ-mental Well-being (CFG)". Twenty-four items were used to measure QL, CEW, and CFG on the scale developed by Quoquab et al. (2019). The measurement of the constructs of environmental knowledge and sustainable consumption behaviour intention was based on the research conducted by Saari et al. (2021). Both constructs included three items each. The environmental influences construct included three items based on the operationalisation applied and validated by Figueroa-García et al. (2018). Nine items for measuring materialism were adopted from the scale of Lindblom et al. (2018) and Ponchio and Aranha (2008). Sustainable consumption promotion involves information about environmental issues, green product promotion, the promotion of recycling, and other external initiatives. To measure PSC, four items from the scale provided by Piligrimiene et al. ˙ (2020) were used. All items were measured on a seven-point Likert scale [where 7 specifies a positive opinion (Strongly Agree/Always) and 1 denotes a negative opinion (Strongly Disagree/Never)]. The measurement items have been listed in Appendix B (Table A2).

The questionnaire was professionally translated into Bulgarian by two bilingual experts from the Department of Language and Specialised Training of Foreign Students at one of the largest Bulgarian universities, and then a standard back translation procedure was used to ensure that the translated content conformed with the original English meaning.

## *3.2. Sample and Data Collection*

The survey design followed a sequence of steps, including a pilot test with 50 respondents aimed to identify problematic items and further improve the survey (Fink 2016). The data were collected with the help of a certified sociological agency operating in Bulgaria. A quota sample was formed for the study of three defined characteristics (gender, age groups, and place of living in administrative-territorial regions in the country). It reproduced the structure of the population in Bulgaria as of 31 December 2021 (in conformity with the data published by the National Statistical Institute in the Republic of Bulgaria). A total of 522 respondents were approached for the questionnaire-based online survey. Observations with missing values and straight lining were deleted, leaving a total sample size of 489. This sample size highly exceeded the recommended minimum sample sizes of 160 and 146 as indicated by the inverse square root and gamma-exponential methods, respectively (Kock and Hadaya 2018).

The respondents' age ranged between 16 and 64 years, with an average age of 40. The age group (based on the age groups formulated by the National Statistical Institute in the Republic of Bulgaria) and the other sample characteristics have been presented in Table 1.


**Table 1.** Socio-demographic profile of the sample.

#### **4. Discussion and Results**

This study employed partial least squares structural equation modelling (PLS-SEM) and the latest software version of SmartPLS–SmartPLS 4 (Ringle et al. 2022) to test the hypothesized relationships. SmartPLS software is widely used in many social science disciplines because of its variety of capabilities and user-friendly features. It can estimate very complex and higher-order models, with a considerably smaller sample size at the same time (Sarstedt et al. 2021).

The analysis pursued the guidelines, procedures, and cut-off values as suggested by Hair et al. (2021). The skewness and kurtosis tests were assessed. The findings indicated that the assumption of normality was violated for some items since the threshold of the absolute skewness value and the absolute kurtosis value exceeded 1.

A two-step process was followed, where the measurement model (outer model) was analysed first, followed by the structural model (inner model). To assess the significance of the path coefficients and the loadings, a bootstrapping method with 5000 resamples was used.

Harman's single-factor test was performed to detect common method bias (CMB) before proceeding to measuring the structural model and the measurement model. The result based on the unrotated principal axis factoring revealed that the first factor explained 34.7% of the total variance, which was less than the critical value of 50% (Fuller et al. 2016). Henceforth, CMB was not at all a concern in the present study.

The model in the present study contained eight first-order reflective constructs and a second-order reflective-reflective construct. The PLS-SEM literature outlines several approaches to the estimation of models containing higher-order constructs, such as the repeated indicators approach and the two-stage approach (embedded and disjoint) (Sarstedt et al. 2019). Since these approaches provide highly similar results when sample sizes are sufficiently large, the disjoint two-stage approach was chosen for the current research. In the first stage of the approach, the model was estimated for reliability and validity with only first-order constructs. After the evaluation of the model, the construct scores for the SCB subconstructs QL, CEW, and CFG were obtained and named QL\_LV, CEW\_LV, and CFG\_LV, respectively. In the second stage, variables QL\_LV, CEW\_LV, and CFG\_LV were used as SCB indicators for the purpose of assessing the hierarchal model. The structural model assessment was created on the grounds of the stage two results.
