Environmental knowledge is defined as the knowledge that integrates know-how related to the functioning and problems of ecosystems, the civic behavior alternatives and the obtention of a greater environmental benefit (Kaiser and Frick, 2002; Kaiser and Fuhrer, 2003; Frick et al., 2004). Subsequently, Mostafa (2007) indicates that the definition of environmental knowledge includes know-how on the environment, the factors that cause its effects, the insights as a whole and the joint responsibility toward sustainability. Additionally, the concept should involve everything from the awareness of environmental problems to their possible solutions (Kollmuss and Agyeman, 2002; Bamberg and Möser, 2007; Zsóka et al., 2013). Based on the contributions of Mostafa (2007), two types of knowledge can be identified:

Abstract (which considers the problems, causes, and solutions of environmental decline).

Thus, environmental knowledge can increase people’s concern, as abstract knowledge, and can cause behavioral changes in individuals, as concrete knowledge (Kollmuss and Agyeman, 2002; Bamberg and Möser, 2007; Mostafa, 2007). In addition, knowledge of the causes and effects of environmental pollution generates an increase in the individual level of awareness, leading to the adoption of a pro-environmental behavior (Onel and Mukherjee, 2016).

From the business, the concepts of Mostafa (2007) can be taken up again. Abstract knowledge in business increases concern for biodiversity and environmental protection in behavioral changes of employees. While concrete knowledge contemplates how business activities, decisions, and activities of employees, owners, and other actors decrease or benefit the environment (Shepherd and Patzelt, 2011; Lülfs and Hahn, 2014; Soto-Acosta et al., 2016; Ploum et al., 2017; Tur-Porcar et al., 2018; Rodríguez-Jasso, 2021).

Taking into account the aforementioned, it becomes particularly relevant to mention that sociodemographic factors, such as gender, lead to a difference in decision-making, and are related to both the concern and the environmental behavior resulting from the individual’s abstract and concrete knowledge (Kollmuss and Agyeman, 2002; Bamberg and Möser, 2007; Mostafa, 2007; Gifford, 2014; Nanggong and Rahmatia, 2018). In this case, Zhou et al. (2014) claim that the gender difference is one of the most fundamental variations between people, and Strapko et al. (2016) emphasize the fact that the social gender roles impact environmental concerns. Specifically, women are characterized by having greater environmental knowledge, attitude, and behaviors (Longhi, 2013; Casaló and Escario, 2018; Rivera-Torres and Garcés-Ayerbe, 2018; Casaló et al., 2019). Sierra-Barón et al. (2018) describe certain specific behaviors that jointly take place in the home, such as waste separation and water consumption.

Perceived behavioral control is a variable that the Theory of Planned Behavior proposes as a predictor of behavior (Ajzen, 1991). Said theory states that behavior is influenced by three types of conditions: the attitude toward the behavior, the subjective norms, and the perceived behavioral control over the behavior (Ajzen, 1991, 2001). Perceived behavioral control involves the availability of opportunities and resources necessary for the behavior, such as time, money, and abilities (Ajzen, 1985, 1991).

Furthermore, perceived behavioral control is based on control related beliefs regarding the elements that may aid or impede the behavior; the control elements include the necessary knowledge and abilities, the presence or absence of time, money or other resources and the participation of people, among others (Ajzen, 2020). If the individual has a sufficient level of real control over his/her actions, it is expected that he/she act in accordance with his/her intentions, when the opportunity arises (Bosnjak et al., 2020).

In the specialized scientific literature, perceived behavioral control has been defined as the individual’s perception regarding his/her ability to carry out the environmental behavior (Karimi and Mohammadimehr, 2022). According to Ajzen (1985), this behavioral control can predict a behavior, as it is more probable that a person will carry out a behavior, if he/she is confident he/she is able to do so. It has been proven that this variable has positive effects on environmental behavior, such as the reduction of food waste (Soorani and Ahmadvand, 2019), green shopping (Mohiuddin et al., 2018; Lavuri et al., 2021), or recycling behavior (Wu et al., 2022).

The concept of employee green behavior refers to any human behavior that does not damage the environment, but in fact improves or preserves it (Kollmuss and Agyeman, 2002; Steg and Vlek, 2009). Following this line of reasoning, Valencia-Ordóñez et al. (2021), when analyzing the link between culture and environmental behavior, explain that behavior is defined as the actions caused by external stimuli and internal motivations that individuals acquire through experiences, if said actions become habitual. In that sense, Anderton and Jack (2011) identify the fact that, when people want to have an eco-friendlier behavior but do not know where to start or do not have the time to make the changes, the workplace represents a path for people’s environmental behavior, toward a more sustainable future.

Applied directly to the business context, when an entity is environmentally sustainable, its financial performance gains considerable benefits (Lourenço et al., 2012; Griffin et al., 2017). The aforementioned is a pattern for many enterprises to focus on building, together with their employees, new measures necessary for attaining organizational sustainability (Linnenluecke and Griffiths, 2010; Nations Global Compact, 2013). Thus, employee green behaviors have been useful in contributing to the creation of the green enterprise (Lu et al., 2017), due to the fact that the behavior of its staff favors the reduction of negative effects on nature (Lülfs and Hahn, 2013). There is empirical evidence that indicates a significant association between green behaviors and workplace satisfaction and organizational commitment (Paillé and Boiral, 2013). This proves that creating a propitious work environment, where employees feel supported by their organization, increases the employees’ willingness to get involved in behaviors that protect nature, as it is claimed that in a workplace context, the employees may exhibit certain voluntary actions if they are backed by their organization (Lavelle et al., 2007; Ones et al., 2010). Nonetheless, Bingham et al. (2013) make it clear that in order to achieve that, the staff must be aware that their organization is seeking environmental sustainability.

Bolzan de Campos (2008) indicates that the employees of certified companies may exercise a green behavior more frequently than the employees of non-environmentally-certified companies, and that this behavior can be explained by collective interest. Employees make efforts to carry out voluntary environmental practices and participate in environmental programs; in addition, some studies have studied how women have environmental behaviors in their workplaces (Safari et al., 2018; Dakhan et al., 2020). Similarly, certain studies indicate the influence of the sustainable voluntary initiatives of a company on the behavior of its employees, where it is established that the activities carried out, as part of these programs, promote a better environmental behavior, improving the respective workplace culture (Zibarras and Ballinger, 2011; Norton et al., 2012).

In order to emphasize the importance of green behavior, Kim et al. (2016) explain it from the perspective of motivations, environmental concerns, and self-efficacy, by exploring these variables among different generations of workers, and claim that environmental concern and self-efficacy are significantly different between generational groups. Nonetheless, there are many factors that impact the adoption of green behaviors in the business sector, such as the regulation of environmental laws, social pressure, and the pressure of environmentalist groups, or simply because this is viewed as a competitive advantage, as it attracts more consumers (Vargas Martínez, 2015).

Within the study of green behavior, it has been noted that perceived behavioral control is an important factor which can explain this variable (Mohiuddin et al., 2018; Fatima et al., 2022; Karimi and Mohammadimehr, 2022; Wu et al., 2022). Additionally, green behavior has been studied and discussed within scientific literature alongside psychographic and demographic elements (Patel et al., 2017). In the specific case of the sociodemographic factors such as age, gender, residence (rural or urban), and identification with one group, among others, has been linked to environmental concern and behavior (Gifford, 2014).

More specifically, it has been identified that women’s commitment to environmental causes is greater than men’s, as the former are more involved in housework (De Gobbi, 2013). Strapko et al. (2016) indicate that women seem to be more concerned with the environment because they have been socialized in care responsibilities. By contrast, men take into account risks or gains when deciding to participate in environmental activities that reflect certain perception or judgment (Eisler et al., 2003). Certainly, it is more probable that men have ethnocentric perspectives and women have ecocentric perspectives with respect to environmental responsibility (Müderrisoglu and Altanlar, 2011). Given this situation, Nanggong and Rahmatia (2018) agree when stating that women are more aware of ecology and of how to act ethically in order to safeguard the environment than men.

Studies on green behavior indicate that women have consistently higher levels of environmental protection and concern than men, as they are more prone to promoting such activities both in developed and emerging countries, and they are more active participants in green consumption, the difference in their green behavior may be due to their variable gender beliefs and roles (Heyl et al., 2013; Calabrese et al., 2016; Patel et al., 2017).

The theory of planned behavior proposed by Ajzen (1991), is defined as the person’s intention to carry out a particular behavior and explains the origin of behavior based on behavioral intention, determined by attitudes toward the environment, perceived behavioral control—which facilitates or hinders the performance of the behavior—and the subjective social norms, all these preceded by behavioral, normative, and control beliefs, respectively. Likewise, he considers individual, social, and informational elements as precursors of beliefs.

This theory differs from the theory of reasoned action by the fact that it adds the variable of the perceived behavioral control (Ajzen, 1991). In turn, the theory of reasoned action: (1) establishes that people act rationally by using available information before acting and (2) establishes that intentions are determined by the attitude toward the specific behavior and by the subjective norms (Fishbein and Ajzen, 1975).

Being one of the most used models in various studies, the theory of planned behavior states that human conduct is voluntary and determined by behavioral intention (Ajzen, 1991; Bosnjak et al., 2020). Thus, social attitudes arise from the interaction between behavioral expectations and their corresponding assessment by each subject, while subjective norm refers to the way in which the subjects receive and interpret what people and groups they consider relevant say about what they should do in relation to the behavior, and the motivation to accommodate these opinions; lastly, perceived behavioral control references the subjects’ beliefs regarding their own ability to carry out a given behavior (Ajzen, 1991). That is why this theory is considered applicable to any voluntary behavior that people in general may exhibit (Ajzen, 2001), a strong link between the intention to perform a particular behavior and its performance (Ajzen, 1991).

Authors such as Yadav and Pathak (2017), Tam (2020), and Ordoñez Abril et al. (2021) agree that this theory has served to explain environmental behavior, taking into account people’s intention when participating in organized pro-environmental actions. Therefore, this theory has been used to improve understanding regarding what kind of intentions motivate employees to carry out green practices (Greaves et al., 2013; Bouarar and Mouloudj, 2021).

Morren and Grinstein (2016) confirm the aforementioned with results of their meta-analysis: the theory of planned behavior has been applied in numerous studies on environmental behavior, specifically in those that seek to analyze the relationship between green attitudes and behaviors. Other empirical findings that credit the approach used in this paper claim that the theory of planned behavior significantly predicts employees’ intentions to carry out environmental behaviors: for example, going to the workplace using an alternative means of transport and making eco-suggestions aimed at the workplace, turning off their computers each time they leave their office for an hour or more, using videoconferences for meetings which otherwise would require traveling and recycling as much waste as possible (Greaves et al., 2013; Yuriev et al., 2018). In turn, Bouarar and Mouloudj (2021) found that the tendency toward behavior, the subjects’ own norms and the environmental knowledge have a positive and statistically significant effect on the employee’s intention to implement green practices.

There are other frameworks that can be used in the study of green behavior such as institutional theory or stakeholder theory. These theories focus on explaining how pressures, whether institutional or from actors, promote behavior (Friedman and Miles, 2002; Chu et al., 2018). Since this research sought to analyze the volitional and non-volitional processes of the women and not the pressures, the theory of planned behavior was used with the difference of testing the relationships in the Mexican context that has not been analyzed.

Various studies on environmental knowledge indicate that it can influence more than just the attitude in a green behavior (Synodinos, 1990; Barreiro Fernández et al., 2002; Fraj and Martínez, 2005). Under this premise, the studies on green behavior find that perceived behavioral control presents a relationship between attitudes and subjective norms (Autio et al., 2001; Joensuu et al., 2013). However, more recent studies such as the one by Salimi (2019) indicate that the positive effect of environmental knowledge is greater on the perceived behavioral control than on the attitude and on the subjective norms, when the intention for green shopping is analyzed. Thus, Torres Canales and Zea Ticona (2021) explain the importance of knowledge in achieving behavior, because if the person has little information and his/her available resources have changed (perceived behavioral control) it is unlikely that the behavior is carried out.

According to Kim et al. (2014), environmental knowledge increases the individual’s perception of his/her control in participating in sustainable volunteering programs. Similarly, Maichum et al. (2016) claim that knowledge influences perceived behavioral control, so that when consumers know about eco labels and certifications, they expand their understanding of the processes and impacts of products, feeling more confident and able to carry out their eco-friendly shopping; in addition, if they consider that they have more resources and time, there will be more willingness and opportunities for them to make the purchase. In other behaviors, such as entrepreneurship, gender has a positive and significant influence on the relationship between subjective norms and perceived behavioral control; that is because society considers entrepreneurship a desirable option for women (Díaz-García and Jiménez-Moreno, 2010), which makes their perceived behavioral control a resource in their entrepreneurial spirit.

That said, in the case of women, it is possible that environmental knowledge increases perceived behavioral control, as posited in the following hypothesis:

The analysis of scientific literature has detected that environmental knowledge is significantly related to green behavior (Azila et al., 2012; Fawehinmi et al., 2020; Zhang et al., 2021; Wu et al., 2022). These findings are in line with the arguments of Lavuri et al. (2021), who claim that young consumers’ environmental knowledge has a positive and statistically significant influence on their purchasing intention. Said positive and statistically significant effect was also found between the consumers’ environmental knowledge and green shopping behavior variables (Azila et al., 2012).

It is important to mention that, in an educational context, environmental knowledge had a positive influence on the green behavior of Iranian students (Karimi et al., 2021). Simultaneously, the environmental knowledge of Chinese students predicted their day-to-day recycling behavior (Wu et al., 2022). Meanwhile, in organizational contexts, findings also indicate that this relationship prevails, with environmental knowledge and environmental knowledge exchange having a positive and statistically significant effect on the green behavior of employees in different organizations (Rosales Palomino, 2017; Fawehinmi et al., 2020; Karmoker et al., 2021; Zhang et al., 2021).

According to Safari et al. (2018), employees’ knowledge regarding environmental issues leads to an increase in their efforts to carry out tasks that improve their company’s environmental performance, even if these efforts are not mandatory. Additionally, the same authors claim that environmental knowledge also influences employees’ desire to make suggestions regarding environmental practices, to participate in environmental programs and to question environmentally harmful practices.

The role of women stands out in the study of green behavior: their beliefs and the gender roles they play have an important part in understanding their green behavior (Patel et al., 2017). There is empirical evidence that allows for the explanation of how women’s environmental knowledge had a significant influence on their having a green behavior in the workplace (Dakhan et al., 2020). Based on the aforementioned, the following hypothesis is posited:

Perceived behavioral control has been linked not only to green shopping behaviors but also to the adoption of green technologies and green behaviors in day-to-day life, among others (Mohiuddin et al., 2018; Fatima et al., 2022; Karimi and Mohammadimehr, 2022; Wu et al., 2022). However, there are few studies that agree with Karimi et al. (2021), who found that the perceived behavioral control of working students had a positive influence on their employee green behavior.

In the specific case of this study, we take into account the findings of Yadav and Pathak (2017), Zierler et al. (2017), Arli et al. (2020), and Fawehinmi et al. (2020, 2022) as basis for analyzing the effect of perceived behavioral control on employee green behavior, because they describe that a higher perceived behavioral control leads to a higher employee green behavior. Hence, the following hypothesis is posited:

Hagger et al. (2022) consider that perceived behavioral control acts as an intervening variable, not just as a variable that determines conducts, because the lack of control may impede the performance of said action. Additionally, it is considered necessary to possess time, money or resources and abilities, for intention to be more probable (Yang-Wallentin et al., 2004; Hagger et al., 2022). Given that the theory of planned behavior expands on the theory of reasoned action, which states that action is rational and informed, perceived behavioral control can be considered a variable that intervenes in the relationships (Fishbein and Ajzen, 1975; Ajzen, 2001; Hagger et al., 2022), this perceived control serving as a variable that mediates the relationship between knowledge and behavior. If the person possesses information and control of his/her available resources, it is more probable that the action will be carried out (Torres Canales and Zea Ticona, 2021).

It was found that knowledge influences environmental behavior (Karimi et al., 2021; Wu et al., 2022), and also that knowledge influences perceived behavioral control (Kim et al., 2014). In addition, perceived behavioral control influences environmental behavior (Yadav and Pathak, 2017; Zierler et al., 2017; Fawehinmi et al., 2020, 2022). When analyzing the intention of green shopping, environmental knowledge has a positive and significant effect on perceived behavioral control, and the latter in turn has a positive and significant effect on the intention of green shopping (Maichum et al., 2016).

In turn, it was proven that through the consumers’ ability to purchase and their resource availability, such as money and time, environmental knowledge leads to the generation of a green purchase, due to the fact that perceived behavioral control plays a mediating role between knowledge and the green shopping behavior (Salimi, 2019). In turn, Karatu and Mat (2015) found that perceived behavioral control has a partial mediating effect between environmental knowledge and the intention of green shopping. Hence, perceived behavioral control can be a mediator of the relationship between knowledge and green behavior, as posited in the following hypothesis:

Environmental knowledge was operationalized and adapted by Fawehinmi et al. (2022) and Gatersleben et al. (2002) and was made up of four items. In order to respond to them, the minimum response value was 1 (not at all) and the maximum value was 5 (to a great extent). In the case of the variable perceived behavioral control, this was operationalized and adapted by Swaim et al. (2014), made up of five items measured using a five-point Likert scale, where the minimum value was 1 (completely disagree) and the maximum value was 5 (completely agree). As for employee green behavior, it was operationalized and adapted by Fawehinmi et al. (2022) and Blok et al. (2015), made up of seven items. Regarding the response options to the items of said construct, the minimum response value was 1 (never) and the maximum value was 5 (always).

The questionnaire was administered to female SME employees in Ensenada, Baja California, Mexico. With regard to the data collection, the questionnaires were applied face-to-face with the participants during the months of February, March, April, May, and June of 2022. In total, 240 valid questionnaires were collected in Ensenada, Baja California, Mexico.

The city of Ensenada is located in the northwest of the Mexican territory, in the northern part of the Baja California peninsula. The Municipal Institute of Research and Planning of Ensenada, based on a diagnosis of the state’s economic fabric, highlights the importance of this region in the primary sector, due to its agricultural, fishing, and aquaculture activities; in the trade sector and the manufacturing industry sector (Instituto Municipal de Investigación y Planeación de Ensenada B. C. (IMIP), 2011). However, the most important weakness in Ensenada is apparently the competitiveness of the companies, it is observed as the lowest in the entire state of Baja California, since there is a phenomenon of pulverization of companies, especially national ones, increasing in number, but decreasing its size (Padilla et al., 2018).

On the other hand, this city has an urban water demand of 920 l/s, but they are only supplies of 745 l/s and by 2030 it is expected that said demand will increase to 266 l/s, so this deficit and future sources of water for the region are uncertain due to low rainfall, over-extraction of groundwater, and saline intrusion of aquifers (Elizondo and Mendoza-Espinosa, 2020).

Secondly, the choice of the female sample is based on the fact that, in Mexico, female labor participation has grown substantially and higher than that for men. In the period between 2005 and 2022, it increased by 4.6% while for men it decreased by 3.3%. In the first quarter of 2022, the female employed population increased 3.7%, adding more than 810,000 new jobs for women (Instituto Mexicano para la Competitividad, 2022), in such a way that their behavior within SMEs has gradually acquired greater importance.

In such a way that the growth of the city will be limited by the lack of water resources, in this sense, to evaluate the level of knowledge of the employees of SMEs, whose participation in the active economic population is increasing and how said knowledge has an effect in practices within companies, it is necessary.

Additionally, the next aspect worth mentioning is the minimum sample size in order to carry out the structural equation modeling by means of partial least squares: having two predictive variables, a level of significance of 0.001, a statistical power of 0.8 and a medium-sized effect, the minimum sample size required is 98 (Cohen, 1988; Nitzl, 2016).

The data were analyzed using the statistical technique of structural equation modeling based on partial least squares (PLS-SEM). The latter is based on the intention to prove and validate an exploratory model and to posit proposals for future links between constructs, without forgetting that PLS-SEM provides concurrent analyses for both structural and measurement models, which leads to more accurate results (Reinartz et al., 2009; Qalati et al., 2022a).

First, we conducted the evaluation of the measurement model, using the consistent PLS algorithm. Being a completely reflective model, the criteria evaluated were: internal consistency, convergent validity, and discriminant validity (Hair et al., 2019). In this case, Table 2 indicates the values of internal consistency and convergent validity.

Evaluation of the Cronbach’s Alpha coefficient and of the composite reliability index establishes that the construct measurement has an appropriate consistency, avoiding possible interpretation biases due to over-or underestimation (Cronbach, 1951; Nunnally and Bernstein, 1994). That said, based on Table 2, we can note that the values obtained for both measurements indicate that the reliability is higher than 0.90 and under 0.95 (Cheah et al., 2019; Hair et al., 2019). Next, we reviewed the indicator loadings and decided to preserve those with loadings greater than 0.708 as they account for more than 50% of the construct variance (Galván-Mendoza and Esquinca-Moreno, 2019; de Rodríguez-Jasso and Sánchez-Limón, 2022). Nonetheless, of the five indicators that registered factorial loads below the cutoff point (<0.708), four of them were eliminated (EN04, EGB06, and EGB07 y PBC05). Item EN01 was preserved, even though it registered a factorial load of 0.523, taking into account Hair et al. (2019:151), who point out that “more than automatically eliminating those indicators with loads under 0.70, researchers should carefully assess the effects that the purging of an item has on the composite reliability and on the construct content validity.”

In order to examine the convergent validity of the measurement model, we took into account a value greater than 0.5 of the variance extracted from the measurement (AVE; Hair et al., 2019; Leyva-Hernández et al., 2021), obtaining satisfactory values (see Table 2). The aforementioned offers certainty that the constructs of this research are unique in their composition. Similarly, we have evaluated the rho_A (ρ_A) values obtained: each one of the constructs showed values high above the cutoff point of 0.70 (Chin, 1998).

In terms of the discriminant validity, we examined the Fornell and Larcker criterion, the Heterotrait-Monotrait Ratio (HTMT), and the cross-loading analysis (Fornell and Larcker, 1981; Barclay et al., 1995; Henseler et al., 2015). In that context, it is necessary to mention that the Fornell and Larcker criterion represents the amount of variance that a construct captures from its indicators, which must be higher than the variance that the construct shares with the rest (Martínez Ávila and Fierro Moreno, 2018). Table 3 shows the formation of a diagonal with the highest values registered in the data run.

The next criterion used to assess the discriminant variance was the Heterotrait-Monotrait Ratio (HTMT). To that end, authors such as Imran et al. (2017) and de Rodríguez-Jasso and Sánchez-Limón (2022) recommend values inferior to 0.90 in order to obtain a satisfactory result and, in turn, Franke and Sarstedt (2019), Hair et al. (2019), and Qalati et al. (2022a) suggest that the value should be less than 0.85. Indeed, Table 4 shows that all values are acceptable and satisfactory, as they are less than 0.85.

Cross-loadings are the last criterion that makes up the discriminant validity analysis of this study. It is necessary to compare the factorial cross-loadings of the indicators that integrate a latent variable with the loads of the indicators of the other latent variables (Barclay et al., 1995; Galván-Mendoza and Esquinca-Moreno, 2019). Thus, the contents of Table 5 show that none of the items has a higher load in another construct, other than the one it measures.

In this point of the analysis, it is considered that the discriminant validity of the constructs included in the modeling we carried out is acceptable. Hence, its measurement is empirically unique.

Similarly, we have assessed the structural model, using the following criteria: the path coefficients (β), the significance values (t value and p value), the effect size (f²), the collinearity of the model, the Stone-Geisser (Q2) values, and the R² coefficient of determination (Benitez et al., 2020; Leyva-Hernández et al., 2021; de Rodríguez Jasso et al., 2022).

This study used the bootstrapping technique consisting of 5,000 samples for 240 cases, generating path coefficients (β) and its significance level. With regard to the p value, it must be less than 0.05 for there to be statistical significance. The t value must be greater than 1.96 (Hair et al., 2017; Hannachi, 2021; Qalati et al., 2022b). The path coefficients (β) obtained are indicated in Table 6 and Figure 1.

When estimating the path coefficients, which represent the hypothetical links between the variables of the described model, one must not forget that, given their standardization, path coefficients may register values within the-1 to +1 range, thus, the value closer to 1, regardless of the sign, reflects the strongest path (Garson, 2016; Hair et al., 2017). Based on Table 6 and Figure 1, the path coefficients obtained are:

Environmental knowledge, β value of 0.495 statistically significant (t value = 6.548 and p value = 0.000) in its relationship to the endogenous variable perceived behavioral control (environmental knowledge → perceived behavioral control).

Subsequently, we have reviewed the size of the f² effects, which represent the degree to which an exogenous construct explains an endogenous one in terms of R². Taking into account the guidelines of Cohen (1988) and Henseler et al. (2009), if the f² value is greater than 0.02, it corresponds to a small effect, if the f² value is greater than 0.15 it corresponds to a moderate effect, and if the f² value is greater than 0.35 it corresponds to a great effect.

Based on the previous paragraph, when evaluating the first f² effect (environmental knowledge → perceived behavioral control), the value 0.364 indicates a moderate effect (being the highest). The second f² effect was 0.061 (environmental knowledge → employee green behavior), characterized as small, and the third and last f² effect registered at 0.259 (perceived behavioral control → employee green behavior), is also considered a moderate effect.

On the other hand, it is extremely important to carry out the evaluation of the collinearity of the modeling performed, using the variance inflation factor (VIF). The VIF allows for recognizing the variables that are highly inter-correlated. Thus, an adequate measurement should have a value between 3 and 3.3 or less, but greater than 0.20 (Diamantopoulos and Siguaw, 2006; Urías et al., 2007; de Rodríguez-Jasso and Sánchez-Limón, 2022). In this case, with the results obtained, it is assumed that there are no multicollinearity problems between the constructs environmental knowledge, perceived behavioral control, and employee green behavior.

The next criterion that should be reported is the Q² values. According to Geisser (1974), Nitzl and Chin (2017), and Fawehinmi et al. (2022), these values help measure the predictive relevance of the observed data of the path model, which is why they must be calculated using the blindfolding resampling technique with a D value of 7. Considering the contributions of Cohen (1992), we took into account the classification of the effect sizes of 0.35, 0.15, and 0.02, respectively, as great, medium, and small effect sizes. This study discovered the fact that the predictive relevance of the variable perceived behavioral control is small (0.147) and, in turn, the variable employee green behavior registered a predictive relevance characterized as medium (0.159).

To finish with the evaluation of the structural model, the values obtained with the R² determination coefficient were examined (see Table 6). This coefficient represents the value of the explanatory power of the model and the predictive power within the sample in question, where an R² of 0.67 is considered substantial, an R² of 0.33 as moderate, and an R² of 0.19 as weak (Cohen, 1988; Galván-Mendoza and Esquinca-Moreno, 2019; Hair et al., 2019; de Rodríguez Jasso et al., 2022).

The results of this study show that the variable perceived behavioral control registered an R² of 0.245. This means that the variable environmental knowledge explains 24.5% of the total of the construct variance of perceived behavioral control, thus being characterized as a weak determination level or explanatory power. By contrast, the variable employee green behavior obtained an R² of 0.370. This means that the variables perceived behavioral control and environmental knowledge explain 30.7% of the total construct variance of employee green behavior. Said value represents a moderate determination value or explanatory value.

Lastly, in order to evaluate the corresponding adjustment of the model, the following measurements were taken into account: standardized root mean square residual (SRMR), unweighted least squares discrepancy (d_ULS), and geodesic discrepancy (d_G; Zaman et al., 2019; Swierczek, 2020; de Rodríguez Jasso et al., 2022; Qalati et al., 2022b). Table 7 presents the results obtained: the value of the SRMR was 0.045 and it was less than the values of the 95 percentiles (0.071) and 99 percentiles (0.075). The d_ULS value was 0.157, less than the 95 percentiles (0.206) and 99 percentiles (0.238). The d_G value was 0.132, also less than the 95 percentiles (0.167) and 99 percentiles (0.186), respectively.

With regard to the obtained value of the SRMR, Hair et al. (2019) explain that in order to obtain an acceptable adjustment, this must be less than 0.08. Similarly, the unweighted least squares discrepancy (d_ULS), and geodesic discrepancy (d_G) must be less than the values of the 95 and 99 bootstrap percentiles of said indicators (Leyva-Hernández et al., 2021; de Rodríguez-Jasso and Sánchez-Limón, 2022). In sum, the model adjustment is acceptable. Next, Figure 1 presents the structural model which indicates the obtained path coefficients and R² values.

In order to assess the mediating effect of the variable perceived behavioral control, two elements were examined: the significance test of the indirect and of the direct effect (Hair et al., 2017; Leyva-Hernández et al., 2021). Given that the value of the direct effect was statistically significant (β = 0.225 and p value = 0.005), we opted to review the value corresponding to the indirect effect (see Table 8). Similarly, statistical significance was also found in the indirect effect (β = 0.183, p = 0.000). Given this situation, a complementary mediation is identified (Nitzl et al., 2016; Li et al., 2019). In this case, this supports H₄ by identifying that environmental knowledge predicts perceived behavioral control which in turn predicts employee green behavior.