A question of norms and control—factors shaping sustainable energy behavior: a study among various university stakeholders

In this study, we investigate the factors influencing energy-saving behavior within a medium-sized university, providing an example of promoting sustainability within organizational settings from an environmental psychology perspective. We employ the Theory of Planned Behavior (TPB) (Ajzen 1991) as a foundational framework to examine how attitudes, personal and subjective norms, perceived behavioral control, and behavioral intention influence behavior. The TPB is a frequently used model to explain environment-related behavior in organizations or in the workplace. More recently, there have been increasing attempts to improve its predictive power by extending it with additional predictors or by combining it with other models (Value Belief Norm Theory, Norm Activation Model, among others) (e.g., Al Zaidi et al. 2023; Ateş 2020; Aziz et al. 2021; Canova and Manganelli 2020; Khalid et al. 2022). Repeatedly, the TPB has been criticized in particular for giving limited consideration to moral and altruistic motivations in pro-environmental behavior and focusing mainly on social influences (Bamberg and Schmidt 2003; Klöckner and Blöbaum 2010; Abrahamse and Steg 2011; Gao et al. 2017).

The TPB emphasizes the role of subjective norms, which represent the social pressures individuals perceive when deciding whether to engage in a specific behavior. These norms are rooted in an individual’s social environment, distinct from personal norms that stem from internalized values and moral standards (Thøgersen 2006; White et al. 2009; Bertoldo and Castro 2016). Subjective norms primarily operate by motivating individuals to avoid punishment for socially undesirable behavior or to receive rewards for norm-compliant conduct (White et al. 2009; Bertoldo and Castro 2016). Social norms can be divided into descriptive and injunctive norms, the former defining the usual behavior in a given social context, while the latter defining what is morally approved or disapproved (Cialdini et al. 1991; Cialdini 2012). The original TPB construct of subjective norm primarily reflects injunctive social norms, with limited consideration for descriptive social norms (Rivis and Sheeran 2003; White et al. 2009).

Social norms, both descriptive and injunctive, are particularly influential when they stem from social groups relevant to the individual, such as close personal relationships or group affiliations (e.g., family, friends, organizations) (Ashforth and Mael 1989; Van Dick 2001). Empirical studies have shown that an individual’s identification with a particular social group positively correlates with the influence of social norms on behavior (Terry et al. 1999; Fielding et al. 2008; White et al. 2009; Nigbur et al. 2010). In addition, a higher level of identification with the organization can be directly positively related to environmentally friendly employee behavior, as demonstrated in contexts like the hotel industry (Shah et al. 2021). Generally speaking, many work environments can be described as public or social spaces, where one’s own behavior becomes visible to the social environment and is therefore more subject to the influence of social expectations or social norms (Littleford et al. 2014; Nye and Hargreaves 2010). Comparing factors influencing energy-related behaviors at home and at work, some more differences can be identified (e.g., Littleford et al. 2014). One of these differences also relates to the individual options for controlling energy consumption behavior. These can be significantly restricted in the workplace (cf. Bentler et al. 2023), e.g. through the collective use of electrical appliances or central/automatic control. In the context of TPB, effects on the variable perceived behavioral control can therefore be expected.

In a departure from the traditional model of TPB, our study introduces the concept of Personal Norm as a significant factor. Unlike social norms, personal norms shape behavior not primarily for their efficiency or avoidance of social sanctions, but rather by aligning actions with internalized values and moral standards, driven by internal consequences such as feelings of guilt or negative self-assessment (Thøgersen 2006). Personal norms represent self-expectations related to specific actions and encapsulate a sense of moral obligation (Schwartz 1977; cited in Thøgersen 2006). They often stem from the internalization of external, particularly injunctive, norms and their compatibility with core internal values (Thøgersen 2006; Bertoldo and Castro 2016). Several studies suggest that personal norms could potentially govern behavior when activated by social norms in specific contexts (Bamberg et al. 2007; Bamberg and Möser 2007; Klöckner 2013), as outlined by Bertoldo and Castro (2016). The significance of Personal Norms in influencing environmental behavior in the workplace has also been explored in studies utilizing variations of the TPB (e.g. Al Zaidi et al. 2023; Ateş 2020; Mouro and Duarte 2021). Despite some controversy surrounding this subject (e.g. Botetzagias et al., 2015), there is emerging evidence suggesting that the moral norm factor might be considered as a construct that bears a high degree of overlap with attitude or even acts as a substitute for it (Chan and Bishop 2013; Kaiser 2006; Heib et al. 2023). In summary, based on existing evidence, the personal norm can be perceived as a potential substitute for attitude, a finding that emerges from various studies investigating its role in shaping behavior.

Based on these theoretical considerations, we derived the following hypotheses:

The hypotheses are summarized in the following research model (Fig. 1).

When transferring the components of the TPB regarding individual behavior described above to the application context of the university type of organization, the specific characteristics of the university and the associated stakeholder roles must be taken into account (Kastner and Matthies 2014b). This refers in particular to the understanding of the diverse stakeholders at the university—i.e., academic staff, administrative/technical staff, and students. The different university groups have specific characteristics in terms of attitudes, norms or behavioral control,furthermore, they vary in their connection to the university’s goals and have distinct roles, responsibilities, and needs. At the same time, they also differ in their formal characteristics such as the duration of their membership at the university, the goals of their membership (work vs. study, science vs. administration), the resulting daily time commitments, and the collegial networks and communication channels (Endrejat et al. 2015). Additionally, the structural and organizational characteristics of the university with its hierarchies and formal responsibilities, power and decision-making options, should be taken into account. Recognizing this diversity among stakeholders is essential for fostering cohesion and sustainability within the university. Additionally, the university as a whole must also be considered as a type of organization and with regard to its options for action (Huber 2023). In addition to the internal structure, the framework is also important here. In principle, public universities in Germany are financed by the federal states. Within this framework, they have their own options with regard to energy, both in terms of generation (e.g. PV, combined heat and power plant) and consumption (e.g. building refurbishment and energy management systems).

The three primary groups within a university—academic staff, administrative/technical staff, and students—differ potentially in their roles and engagement concerning sustainability. With regard to energy use, they differ very specifically in their ability to influence. While students as users have little ability to act other than light switches and closing windows, those of university employees range from daily usage behavior of various devices to strategic consumption decisions in efficiency measures such as PV, efficient office technology or thermal insulation. In this context, Zacher et al. (2023) offer a classification of EGB along three continuous dimensions. The first dimension concerns the role in the organization, with a distinction being made between “in the role” (i.e. as part of the core tasks of the employees) and “outside the role” (i.e. as an organizational citizen (i. d., p. 7)). Looking at the role, it becomes evident that there is a conceptual connection with the level of identification with the organization (Blader et al. 2017). The second dimension they describe refers to the effects of EGB, which can be direct or indirect. While the direct effects in the context of a university result from the respective behavior, e.g. efficient or wasteful use of energy, the indirect dimension refers to impacts on colleagues, but also on superiors and more distant members of the organization (e.g. people working in other departments or faculties). Finally, the third dimension deals with the intensity of the behavior in terms of effort and risk. This dimension is strongly linked to the options for action of the respective groups and accordingly the behavior control.

Following these considerations, brief characteristics emerge for the three main groups of academic staff, administrative-technical staff and students with regard to their connection with and their role in the university organization:

Looking at the stakeholder differentiation, it is fair to state for all three primary groups that they are no homogeneous systems, but show differences within the groups themselves. Among academic staff, the internal culture can differ depending on the subject area (e.g. natural sciences or humanities), while for administrative staff, everyday working life and opportunities for behavior depend on whether they work in an office or a laboratory (Endrejat et al. 2015). As far as the student subgroup is concerned, studies show that more female students are involved in student organizations than male students (Filho et al. 2023).

Several general hypotheses on energy-saving behavior and correlations with relevant variables in a university environment were formulated above. However, if the three groups mentioned are taken into account, there could be differences in the correlations. Concerning the factors influencing energy-saving intention (Hypothesis 1), distinct patterns are anticipated among these groups. For academic staff, it is expected that their energy-saving intentions might be moderately influenced by their attitude or personal norms, as their focus primarily revolves around research and academic pursuits (e.g., emphasis on advancing knowledge). Social norms, both descriptive and injunctive, might hold a weaker sway due to their less direct involvement in social sustainability norms. However, perceived behavioral control is deemed crucial as specialized areas such as laboratories might pose restrictions on their ability to control energy consumption (e.g., limited control over specific equipment or facilities). In contrast, administrative/technical staff are envisaged to exhibit a different pattern. Their energy-saving intentions might be significantly influenced by their attitude or personal norms, given their pivotal role in implementing sustainability measures within the university infrastructure (e.g., emphasis on integrating sustainability into administrative processes). Furthermore, both descriptive and injunctive social norms might strongly impact their intentions, as they are actively engaged in sustainability initiatives across the campus. Perceived behavioral control is anticipated to be crucial, given their direct influence over infrastructure and administrative processes that affect energy usage (e.g., direct control over facilities or energy-efficient systems). Students, in their transient tenure within the university, are expected to show moderate influences of attitude or personal norms on their energy-saving intentions, alongside moderate impacts of social norms, considering their participation in environmental programs or campus initiatives. Their perceived behavioral control is projected to be significant, considering potential limitations in their control over university resources and facilities during their study period (e.g., limited control over dormitory energy usage or communal spaces).

Thus, additionally to the general hypotheses above we addressed the organizational type university with its specific context and potential differences between the main stakeholder groups with the following hypotheses

The study was part of a 5-years project which aimed at optimizing property-wide energy consumption at a German university campus and ran from May 1st, 2012, to April 30th, 2017. Within this project’s framework, a series of initiatives were undertaken, including comprehensive university-wide awareness and information campaigns.

These multifaceted campaigns encompassed various activities such as employee training sessions in selected departments, open house events, the distribution of informational materials provided by the German Energy Agency (dena), workshops, and activating Carrot Mob events on central places of the university. The overarching objective of these initiatives was to foster awareness and disseminate information about effective energy-saving measures and practices across the university campus. This context provided the backdrop against which the scientific study was conducted, aiming to investigate and contribute insights into energy-saving behaviors within this dynamic and informed university setting.

The data collection via online questionnaire took place in November 2013 (t1) and April/May 2016 (t2). The population addressed by this study consisted of all members of the university, i.e., all staff members (approx. 2800 at t1 and t2) and all students (approx. 18,300 at t1 and 16,500 at t2) which were contacted via a central email distribution list, informed about the study and given link access to the questionnaire. Before starting the survey, the questionnaire was approved by both the scientific and as well as by the non-scientific staff council of the university. The main distribution channel was the general email distribution list of the university. Thus, both employees (academic and administrative/technical staff) and students could be reached with one email. After the initial mail, there was 4 weeks later a reminder via mail. In addition, a link was placed on the university’s Facebook page to better reach students. Furthermore, the survey was advertised on information screens, for example in the cafeteria. The overall response rate was 9% (students 6%, staff members 26%) at T1 and 8% (students 4%, staff members 22%)

To assess the constructs of interest for the study, we used already established scales from other publications as far as possible. This has the advantage that these instruments have already proven to be appropriate in other studies, e.g., regarding their factorial structure and reliability. In other cases some items had to be constructed for this study to answer the study questions. In the following, we present an overview of the instruments used. The items were rated on a five-point Likert-type rating scale ranging from 1 = “do not agree at all” to 5 = “fully agree” except for energy saving behavior which was assessed at a five-point frequency scale from 1 = “never” to 5 = “always”.

The study uses three independent samples for the analyses: 1. the cross-sectional t1 sample (participants only at t1), 2. the cross-sectional t2 sample (participants only at t2), and 3. the longitudinal sample (participants who took part at t1 and t2).

For statistical analyses we used JASP (Version 0.17.3; JASP Team 2023) for simple statistical analyses (e.g., descriptive statistics). For advanced analyses (e.g., structural equation modeling), we used R (R Core Team 2018) with the lavaan package (version 0.6‑6, Rosseel 2012).

The descriptives for both the t1 and the t2 sample are displayed in Table 1. In both the t1 and t2 sample the scale statistics as well as the intercorrelations are quite similar. The descriptives and correlations of the included scales from the longitudinal sample are provided in the annex A1.

Two independent multigroup SEM (“full information maximum likelihood”) were calculated according to the model in Fig. 1, one for the T1 and one for the T2 bsample. An acceptable model fit was found for the t1 subsample (X² = 3330.238, df = 1440, p < 0.001, CFI = 0.91, RMSEA = 0.05, SRMR = 0.07) as well as for the t2 subsample (X² = 2872.129, df = 1440, p < 0.001, CFI = 0.91, RMSEA = 0.05, SRMR = 0.07).

In general, both independent multi-group SEMs showed very similar results, which indicates the validity of these results. Concerning the hypotheses, the effect of descriptive social norm on personal norm was only significant in the group of students at both times of measurement but contrary to the assumption (H1a) with a negative weight (t1: β = −0.129, p < 0.01; t2: β = −0.133, p < 0.05). Injunctive social norm predicted personal norm significantly as assumed in Hypothesis H1b in all three subgroups and at both times of measurement (values ranged from β = 0.168, p < 0.05, administrative/technical staff at t1 to β = 0.386, p < 0.001, students at t1). The energy saving intention was significantly predicted by personal norm (H2a) in all three groups and over both times of measurement (values ranged from β = 0.346, p < 0.001, academic staff at t1 to β = 0.549, p < 0.001, students at t2). Injunctive social norm had only one significant effect on energy saving intention (H2b) at t1 in the group of students (β = 0.073, p < 0.05), descriptive social norm had in no group at no time of measurement any significant effects on energy saving intention (H2c). In line with the assumptions the effect perceived behavioral control on energy saving intention (H2d) was significant in all three groups at both times of measurement ranging from β = 0.276 (p < 0.001; academic personnel at t2) to β = 0.424 (p < 0.001; academic personnel at t1). The assumed effect of energy saving intention on energy saving behavior (H3a) was significant in all three subgroups and at both times of measurement (values ranged from β = 0.340, p < 0.001, students at t1 to β = 0.660 p < 0.001, academic staff at t1). The effect of perceived behavior control on energy saving behavior was only significant for the group of students at t1 (β = 0.137, p < 0.001). The effect of identification with the organization on energy saving intention (H4) was only significant in the group of academic staff at t1 (β = 0.158, p < 0.01).

The SEM subgroup models accounted for between 27.3% (administrative/technical staff at t2) and 46.0% (students at t2) of the variance of energy saving intention and between 14.8% (administrative/technical staff at t2) and 39.4% (academic staff at t1) of the variance of energy saving behavior. The detailed results can be found in Table 2.

For the longitudinal SEM the longitudinal sample was used. Various predictors including personal norm/attitude, injunctive and descriptive social norms, perceived behavioral control, and identification were measured at t1, while intention and behavior were assessed at t2. The model fit for the longitudinal sample initially showed an acceptable fit (X² = 843.717, df = 480, p < 0.001, CFI = 0.86, RMSEA = 0.06, SRMR = 0.08). A slight enhancement in model fit was observed with three adjustments, allowing for the correlation of errors among items within the same scale (X² = 781.733, df = 477, p < 0.001, CFI = 0.88, RMSEA = 0.06, SRMR = 0.08). However, due to limitations in interpretability caused by these adjustments and the marginal improvement in model fit, we chose not to include these modifications in our final calculations.

The outcomes of this analysis largely resemble those observed in the cross-sectional multi-group SEM (see Fig. 2). Notably, there was a significant negative effect of descriptive social norm on personal norm (β = −0.438, p < 0.001), which diverged from previous findings. Moreover, the effect of perceived behavioral control on energy-saving behavior was found to be non-significant (β = 0.026, p > 0.05).

The study’s findings shed light on the significance of adopting a subgroup-oriented approach when investigating green behavior within public organizations. This approach proves essential as merely focusing on employees might overlook critical nuances. Just as our research identified distinct groups within universities exhibiting varying behaviors, similar differentiation might exist in other public organizations, such as hospitals (cf. Hildebrand et al. 2022). These settings involve a diverse array of stakeholders, ranging from patients to visitors and service providers, alongside medical and nursing staff, each requiring tailored interventions.

Furthermore, the limited significance of the descriptive social norm for energy-saving behavior can prompt further exploration into organizational-specific effects. The university system comprises diverse groups of actors, internally differentiated into various units like faculties, institutes, and departments, each maintaining a certain degree of autonomy. This autonomy fosters distinct social identities, potentially constraining both the expression and perception of social norms. This aspect necessitates further investigation, emphasizing the need for systematic comparisons with different organizational types. For instance, comparing the dynamics within university administrations to similarly sized private-sector organizations could offer insights. This pursuit aligns with the call for more comprehensive analyses of social identity theory within organizational research (Blader et al. 2017).

Lastly, the study also confirms the importance of personal norm as a significant variable in explanatory models based on the theory of planned behavior. In this respect, it argues for an extension of the theory. This is because, although the Theory of Planned Behavior is one of the most commonly used theories in environmental research to explain individual pro-environmental behavior and intentions, it explains less than half of the variance in intentions and behavior (Yuriev et al. 2020). Other models such as the Value-Belief-Norm (VBN) model have different theoretical assumptions for the same outcomes, which has led to an unsatisfactory coexistence of different theories (Bamberg et al. 2007). In line with the criticism and previous findings, this study confirms that norms appear to play a greater role in TPB than originally conceptualized. This study complements the rather small number of studies to date that examine the interaction of the three different norms together (personal, injunctive, descriptive) to explain environmental intentions or behavior (cf. Niemiec et al. 2020). In this respect, the results also argue for a more integrative theory of green intention and behavior that integrates important assumptions from TPB and VBN, for instance.

The importance of the personal norm for the intention to save energy—found for all three subgroups and also in the longitudinal section—makes it clear that interventions in universities should primarily address the personal norm. This can be achieved through specific communication methods that address aspects that are important for the personal norm, e.g. background knowledge, attention to the topic and association with environmentally relevant values as well as emotional involvement and personal references and options for action.

In addition, at the organizational level, participatory interventions can help to increase personal norms, as they lead to an internalization of environmental values into the self-concept (Endrejat and Kauffeld 2018). In addition, participatory approaches not only increase perceived self-efficacy, but can also have positive effects on social norms through group experience and identification with the organization. Accordingly, engagement approaches are important in order to utilize the social potential in addition to technical automation for energy savings (Bull and Janda 2018).

Extrinsic incentives (e.g., financial) would probably be rather counterproductive in that case and might even undermine the intention and thus the concrete behavior (e.g., Van der Linden 2015). An exception could be if the financial incentives are not intended directly for the individual, but e.g. could be used collectively for teaching; thus, positive or reinforcing effects would be seen especially in the subgroup of students. In line with that a study in a university context found that collective goals such as reducing costs for the department can indeed be an incentive to adopt energy-saving behavior (Endrejat et al. 2017).

With regard to social norms a prerequisite of personal norms, the use of prompts (written or pictographic cues) can be a way of conveying injunctive social norms, especially in the case of easy-to-perform behavior and a fundamentally positive evaluation of the action or its goal (e.g. Abrahamse and Matthies 2013). Furthermore, (injunctive) social norms should also be made salient for example by influential people within a department/work group or division (cf. Robertson and Barling 2013). The combined use of injunctive and descriptive norms is considered particularly promising (Cialdini 2012). In this context, another way to support the social norm is to create appropriate targets and structures at university level in order to institutionalize the social norm to a certain extent and clearly emphasize its binding nature. In addition to the definition of concrete savings targets, the establishment of a staff unit for energy management, for example, is also an important strategy which, together with communication measures, can increase the awareness of university stakeholders for the topic of energy. An energy manager creates a visible and competent contact person for advice, which in turn can support efficient usage behavior

It is therefore advisable to communicate to students and staff throughout the university that energy-efficient action is a matter of course for the University, at all organizational levels. For example, multipliers could be trained to set an example of appropriate behavior in their social environment at the University (cf. Carrico and Riemer 2011). However, more cost-intensive structural and technical measures would also be required to make it clear that the university is committed to creating structural conditions for greater energy efficiency (cf. Whittle and Jones 2013).

Another recommendation is that the interventions should specifically address the options for action of the individual groups, as these differ greatly. This is all the more important because, in addition to motivation, knowledge about the impact of one’s own actions also plays an important role in environmentally relevant behavior (Kastner and Matthies 2014a). Furthermore, as the results revealed lower perceived behavioral control among students, this can be seen as an important indication to implement participatory measures in order to actively involve students in the organizational transformation process towards sustainability.

The presented study provides valuable insights on the factors being relevant for explaining energy related behavior in organizations, however, we are aware that there are some limitations that should be taken into account. First of all, the usual problems with self-reports in questionnaire surveys should be mentioned. For the behavioral variable in particular, this aspect could be addressed by recording actual energy use behavior in further studies. The described interventions (3.1.) have not been systematically addressed in the sense of a direct and on site evaluation. We therefore assume that they might have an impact on the perception and behavior between the two points of measurement, but we have no systematic data as further indicators.

Within this study, the main focus was on differences between the three primary stakeholder groups: students, academic and non-academic staff in order to reflect the heterogeneity of university stakeholders. We are aware that these three groups are themselves heterogeneous and have different characteristics that could be relevant for the respective energy use behavior. We did not address in particular the in-group differences. Regarding the behavior of the scientific and administrative staff, we did not differentiate in detail the position within the group (e.g. management function, professorship) or the specific office situation as a potential factor. Xu et al. (2020) report that perceived control over energy-saving and perceived ease of access to building control features had no direct impacts on energy-saving behaviors in single-person offices, while they had impacts on energy-saving behaviors in shared offices. This indicates that for a deeper understanding the office situation should be addressed as one relevant factor of an in-group differentiation. Another example of in-group analyses are the differences in perceptions of environmental concerns between the different disciplines of student groups such as natural sciences or humanities as well as the current study level; as the participants reach higher academic levels, they are more likely to participate in climate change activities (Filho et al. 2023).

These specific differentiations should be investigated in further studies, for which this paper provides a good basis.

Lastly, even though we found relatively comparable results in this field study with three independent samples, which indicates a certain validity of the findings, future studies should investigate the relations of these variables further. Other, better established instruments should be used for this purpose. In addition to the longitudinal sample used here, rigorous methodological designs, e.g. intervention studies in a control group design, would be a good next step.