Energy-Saving Behaviours in Workplaces: Application of an Extended Model of the Theory of Planned Behaviour

The proliferation of energy-consuming technologies (e.g., climate-control systems, personal computers) has resulted in an escalation in the use of energy, which has had a negative effect on the environment. This is reflected in the increased interest in environmental protection that can be achieved not only by using technologies that can improve energy efficiency but also by stopping the excessive consumption of natural resources and by practicing voluntarily individual behaviours that preserve the environment for future generations. To save energy, structural changes such as retrofitting buildings with energy-efficient characteristics are important starting points; however, they may not be enough to reduce energy consumption due to the ‘rebound effect’, i.e., they can even actually induce an increase in energy consumption. Therefore, motivating individuals to save energy may make an important contribution to reducing energy use.

Individual energy-saving behaviour has become an important research topic, but most studies have focused primarily on household energy-saving measures, and, to the best of our knowledge, little research has addressed the issue of individual energy-saving behaviour in workplaces (see the meta-analysis by Yuriev, Dahmen, Paillé, Boiral, & Guillaumie, 2020). The differences between domestic and organisational settings yield results at the residential level that are not immediately applicable to workplaces. In fact, these settings can make it easier or harder to perform pro-environmental behaviours (Maki & Rothman, 2017). For example, in a household setting, individual control is likely to be much higher than in an organisational setting such as an office, where behaviours may be shaped by the physical context (e.g., access to equipment control), by the social context (e.g., expectations and norms of other people), and by the organisational context (e.g., organisational culture and structure). Moreover, in the case of energy-saving behaviours, the costs are experienced more directly in the private sphere, while employees benefit only indirectly from the financial advantages of energy saving in workplaces (Littleford, Ryley, & Firth, 2014; Lülfs & Hahn, 2014). For example, a recent study by Littleford et al. (2014) found that people perform some, but not all, energy-saving behaviours consistently at home and work, but the study did not obtain any evidence to support the 'spillover' effect across workplace and household settings. Thus, to improve the efficacy of technical measures and organisational energy-saving policies, it is necessary to better understand employees’ opinions and behaviours. In fact, the impact of one individual’s energy-use decisions might seem small, but in the aggregate this impact will be significant. The present study aimed to explore the socio-psychological factors influencing two specific energy-saving behaviours in the workplace. The reference theoretical model was the theory of planned behaviour (TPB; Ajzen, 1991). The TPB model was extended by considering an additional variable, habit, which is not included in the original theory.

The TPB is a social-psychological model of behavioural decision-making (Ajzen, 1991). Briefly, this model predicts that the stronger the intention to engage in a behaviour, the more likely the behaviour will be performed. Three constructs determine behavioural intention: attitude towards the behaviour (i.e., the favourability a specific action has for an individual); subjective norm (i.e., the perception of expectations of relevant people that a specific action should be performed); and the perception of how easy or difficult it would be to perform the behaviour in a particular context (perceived behavioural control [PBC]). PBC is a construct that captures to what degree individuals believe they have control over their behaviours, including ability, time, skills, and opportunities to perform the behaviours. It is held to predict behaviour directly, along with intention, especially when the behaviour is not under complete volitional control. The TPB has been applied successfully in a wide range of fields, including pro-environmental behaviours (Klöckner, 2013).

The original TPB model has been extended in two main ways: a) the constructs (i.e., attitude towards the behaviour) have been decomposed in two or more dimensions and b) several factors have been explored as additional predictors or moderators of relationships in the TPB. Regarding attitude, two components have been distinguished: cognitive or instrumental and affective or experiential. The cognitive component reflects the act’s perceived instrumentality (i.e., its anticipated positive or negative consequences, perceived benefits and costs of the behaviour), whereas the affective component reflects the positive or negative experiences with the behaviour, as well as any emotion-based judgements about the behaviour (Fishbein & Ajzen, 2010). There is strong evidence that affective attitudes are more powerful predictors of a variety of health behaviours, especially those related to food choice (Blanchard et al., 2009; Canova, Bobbio, & Manganelli, 2020). As far as we know, the two components of attitude have received very little attention in research regarding environmental behaviours (e.g., Rhodes et al., 2015).

Regarding the second area of TPB expansion, i.e., integrating additional psychological factors and variables into the model to improve its predictive power, the most used variables have been past behaviour, personal or moral norms, self-identity, habit, and anticipated emotions.

The TPB has been used to examine pro-environmental behaviours and energy saving in household contexts, and it has been shown to be more successful in predicting behaviours, compared to other variables such as demographic (Abrahamse & Steg, 2009). In his meta-analysis, Klöckner (2013) noted that roughly 40% of all papers published in the environmental-psychological field had used the TPB as their basic theoretical framework, and analysing 56 different data sets with a variety of environmental behaviours showed that intentions, PBC, and habits were identified as direct predictors of behaviour. The strongest predictor was intentions followed by habit, while PBC had a weaker effect on behaviour. Attitudes, PBC, and personal and social norms (in descending order of their impact size) predicted intention.

In recent years, there has been an increasing interest in pro-environmental behaviours in workplaces, and some studies have applied the original or extended TPB models to predict energy-saving behaviours in this setting. Some research have considered, as target behaviours, specific pro-environmental behaviours, such as ‘switching off computer whenever leaving desk’, ‘using video-conferencing for any meeting’, ‘recycling as much waste produced at work as possible’ (Greaves, Zibarras, & Stride, 2013), and ‘printing smaller’, ‘not printing e-mails’, ‘switching off lights’, and ‘switching off monitors’ (Lo, Peters, van Breukelen, & Kok, 2014). In other research, composite indexes derived from several different pro-environmental behaviours were used (Blok, Wesselink, Studynka, & Kemp, 2015; Dixon, Deline, McComas, Chambliss, & Hoffman, 2015; Wesselink, Blok, & Ringersma, 2017), while other studies used a generic target behaviour: ‘energy saving in my company’ (Gao, Wang, Li, & Li, 2017; Zhang, Wang, & Zhou, 2014). In all the studies mentioned above, the theoretical model was the TPB, extended with additional predictors: personal or moral norms and descriptive norms (Gao et al., 2017; Lo et al., 2014); habit (Lo et al., 2014); organisational electricity-saving climate (Zhang et al., 2014); and perceived leadership support and exemplary behaviours (Blok et al., 2015; Wesselink et al., 2017). Only Greaves et al. (2013) explored further the specific antecedent beliefs of attitude (behavioural beliefs), subjective norms (normative beliefs), and PBC (control beliefs), with their results showing that most antecedent beliefs were significant in influencing behavioural intentions mediated via the TPB constructs. The quotas of explained variance in intentions ranged between 74% for the behaviour of ‘switching off monitors’ (Lo et al., 2014) and 22% for a composite index of several pro-environmental behaviours (Wesselink et al., 2017). Generally, attitude was associated with intentions, but regarding subjective norms and PBC, the results were less consistent, and the associations with intentions could range from very weak to strong.

Some limitations of TPB-based research described above have been highlighted (Yuriev et al., 2020). Firstly, most studies focused only on intention (e.g., Gao et al., 2017; Greaves et al, 2013; Zhang et al., 2014); further, those that focused on behaviour considered past behaviour as a proxy for future behaviour (e.g., Lo et al., 2014). However, when intention and behaviour are measured contemporaneously, a consistency bias is likely to inflate the correspondence between them (Hausenblas, Downs, Giacobbi, Tuccitto, & Cook, 2008). Furthermore, while some studies explored specific behaviours (e.g., ‘switching off lights’ or ‘not printing e-mails’), others considered composite indexes derived from many different pro-environmental behaviours, and yet others focused on a generic target behaviour such as ‘energy saving’. This imprecision in the definition of behaviours studied is problematic as the TPB seems unsuitable for exploring non-specific actions (Ajzen, 1991; Yuriev et al. 2020). Finally, the measures of TPB constructs did not follow the guidelines indicated by Fishbein and Ajzen (2010). All these discrepancies making it very difficult to compare results.

In our study, we considered habit as an additional predictor in the TPB model. Habits are viewed as playing a fundamental role in generating behaviour, and the term ‘habit’ is used widely in ordinary parlance to refer to frequent and persistent behaviour. Recently, Gardner (2015) defined habit as "a process by which a stimulus automatically generates an impulse towards action, based on learned stimulus-response associations" (p. 280). This definition includes cue-dependence, automaticity, and conditioned stimulus-response associations that depict habitual behaviour which, in turn, may be defined as "any action, or sequence of actions, that is controlled by habit" (Gardner, 2015, p. 282). Similarly, Hagger (2019) defined habit "as a specific action or behavioural tendency that is enacted with little conscious awareness or reflection, in response to a specific set of associated conditions or contextual cues" (p. 119). Sometimes, habit has been equated with the frequency of past behaviour, but there are good reasons for not accepting such analogy (Klöckner & Matthies, 2004). Past behaviour is the pattern of behaviour shown prior to the actual behaviour. It includes habitual components as well as intentional behaviour; it also includes repeated actions in addition to actions shown only once or occasionally. In contrast, a habit is the result of association between cues and behavioural patterns; this association is learnt by repeating the same behaviour under the same circumstances over and over again. Most scholars in defining habit added references to automaticity and to reduced deliberate thinking.

Verplanken and Aarts (1999) argue that habit should be introduced in the TPB as a further predictor and/or moderator of the intention-behaviour relationship. Most studies that have examined additive and interactive effects of habit are related to the domains of nutrition and physical activity (e.g., Gardner, de Bruijn, & Lally, 2011; Menozzi, Sogari, & Mora, 2017). In many cases, the results showed that habit explained additional quotas of behaviour variance and it moderated the relationship between intention and behaviour, so that intention had a reduced impact on behaviour when habit was strong. However, some studies failed to find evidence for the effect of ‘habit × intention’ interaction, in relation to physical activity (e.g., Murtagh, Rowe, Elliott, McMinn, & Nelson, 2012; Rhodes, de Bruijn, & Matheson, 2010).

Few studies have examined the effect of habit on relationships between the other constructs of the TPB (attitude, subjective norm, and PBC) and intention. In the study of de Bruijn et al. (2007), a multigroup analysis revealed that the impact of affective and cognitive attitudes and of subjective norm on intention increased when habit was low, while the impact of PBC increased when habit was high. Finally, Menozzi et al. (2017), in a study about healthy-eating behaviour, found that attitude and subjective norm had stronger impact on intention when habit was low.

In the ecological field, Klöckner and Matthies (2004) found that habit increased the amount of explained variance in behaviour and also weakened the association between personal norms and behaviour. Interestingly, despite also being recognised as an important determinant of pro-environmental behaviour, habit so far had mostly been neglected in research on energy-saving behaviours. A recent study by Wang, Lin, and Li (2018) analysed a TPB model extended with non-cognitive (personal moral norm and habit) and emotional factors (positive anticipated emotions). The target behaviour was ‘to save electricity in my home’. The model explained 26% of the variance in intention to save electricity and 30% of the variance in electricity-saving behaviour. Personal moral norm, attitude, PBC, habit, and positive anticipated emotion were positively and significantly associated with residents' intention to save electricity; habit also was associated with electricity-saving behaviour.

To the best of our knowledge, habit has not received extensive attention in the literature on pro-environmental behaviour in workplaces (for notable exceptions, see Lo et al., 2014). In fact, Lo et al. (2014) introduced habit in their study on organisational contexts. The examined behaviours were: ‘printing smaller’ and ‘not printing e-mails’ (printing behaviours), ‘switching off lights’ and ‘switching off monitors’ (switching behaviours). The habit was measured with two items selected from Verplanken's index (Self-Report Habit Index; Verplanken & Orbell, 2003). The hypothesised model was a TPB model extended with habit and personal norm. The explained variance of intention ranged from 50% for ‘not printing e-mails’ to 69% for ‘printing smaller’. With respect to behaviour, the quotas of explained variance ranged from 53% for ‘switching off lights’ to 74% for ‘printing smaller’ and ‘switching off monitors’. The examination of path coefficients showed that habit was a significant predictor of intention for all behaviours and that intention was the strongest direct predictor of printing behaviours, while habit was the strongest direct predictor of switching behaviours. Attitude had the strongest effect on intention for all behaviours, while the effects of subjective norm, personal norm, and perceived control were less consistent across behaviours. A limitation of this study was represented by the use of self-reported behaviour with reference to the past month as a proxy for future behaviour.

Method

Results

Discussion

The findings of this study sustained the validity of the TPB in predicting energy-saving behaviours in workplaces, providing further evidence of the efficacy of this model. The results showed that intention was the only direct predictor of the two behaviours. In turn, intention was associated with cognitive attitude, subjective norm and PBC. Hence, the three hypotheses H1, H2, and H4 were supported. Intention performs as an integrative variable gathering the impact of cognitive attitude, subjective norm, and PBC. Hypothesis H3 was partially supported. In fact, PBC was the strongest predictor of intention to switch off both lights and electronic devices, but it is was not directly associated with behaviours. In the literature, the overall evidence concerning the efficacy of PBC in directly predicting behaviour is not consistent. Additionally, it has been suggested that PBC represents not only perceived physical control but also perceived difficulty or even social appropriateness, which might clarify why PBC is an important construct to consider when predicting people's intentions to conserve energy in public settings. Some studies suggest that perceived control is more important in office settings (e.g., Littleford et al., 2014) or in other contexts outside of the home (e.g., Maki & Rothman, 2017) and is perhaps less relevant in one's own home. However, the finding that intention predicted behaviour but PBC did not suggests that, for the energy-saving behaviours considered in this study, while PBC serves to shape intention—even after controlling for the effects of habit—it does not predict behaviours that seem to be under individual volitional control.

The core variables of TPB explained high quotas of variance intention (56.4% and 64.6%). These percentages of accounted variance, similar to those reported by Greaves et al. (2013) and Lo et al. (2014), might be because in our research, we focused on highly specific behaviours that may not be open to alternative interpretations. Therefore, the possibility of incompatibility between the person's representation of the attitude target and the actual target, towards which the attitude and intention are directed, was very low. Affective and cognitive attitudes towards the behaviours were favourable, but only the cognitive component of attitude was associated with intention. Cognitive attitude reflects the anticipated positive or negative consequences and perceived benefits and costs of the behaviour. It is interesting to note the differences in effects that the two components of attitude have, compared with the behaviours of food choice and consumption. For these behaviours, affective attitude, i.e., the pleasantness of foods, emerged usually as a predictor of intention to consume a certain type of food, as consumption of most liked food usually leads to enjoyment (Canova et al., 2020). The two components have not been studied specifically in the energy-saving domain, but the cognitive component has considerable similarity with moral norm, given the environmental impact of energy-saving behaviours. Moreover, our results, in line with those of Rhodes et al. (2015), support the position that the two components of attitude have different relationships with intention, even in the context of pro-environmental behaviours.

Adding habit increased the portion of variance explained in intentions and behaviours. In the case of behaviours, the increase, although significant, was lower. The quotas of additional variance explained in intentions were similar to those reported in the literature. However, the relatively low explained variance of behaviours, when habit was also considered, indicated that other motivational, volitional, and/or environmental factors need to be considered for predicting energy-saving behaviours. The intention-behaviour gap, that is, the difference between a participant’s intention to act in a certain way and the actual behaviour, is very common in TPB-based studies (Yuriev et al., 2020). Some reasons for this gap have been highlighted: a) events that occurred between assessment of intentions and assessment of behaviours may have produced changes in intention, b) unanticipated obstacles may have prevented the individuals from carrying out their intentions, and c) measures of actual behaviour can be fallible both with respect to reliability and with respect to construct validity. Hence, validate measures of pro-environmental behaviours would allow researchers to identify and evaluate actual behaviours with more precision (de Leeuw, Valois, Ajzen, & Schmidt, 2015). Habit was a strong predictor of intention, but it was not directly associated with behaviours, and its effect was mediated by intention in the case of switching off lights. Thus, the hypothesis H5 was partially supported. Our results on the relationship between habit and behaviour are different from those of other studies (e.g., Lo et al., 2014); however, we must bear in mind that in our research, the behaviours were surveyed at a later stage than the one in which habit and intentions were detected. Finally, habit did not moderate the intention-behaviour relationship. Therefore our hypothesis H6 was not supported. Our findings—in line with Murtagh et al. (2012), do not support the behaviourist view that habit diminishes the effects of rational decision-making (e.g., intention) on behaviour—suggest that motivational and habitual processes have complementary but independent effects on the two switching behaviours. Nevertheless, habit intervened to modify the strength of some relationships between predictors and intentions: a) if individuals had low habit in switching off lights, the intention to do it was more likely if they perceived the behaviour under their control; b) still, if individuals had low habit of completely switching off their electronic devices, the intention to do it was more likely if they developed a strong cognitive attitude, i.e., if they positively evaluated the consequences of their behaviour; c) finally, if individuals had a poor habit of switching off electronic devices, social pressure had no effect on the formation of intention.

This study has a few limitations that should be considered in generalising its results and in choosing directions for future research and interventions. First, we interviewed a convenience sample of workers from different organisations; therefore, the results may not be generalisable. Moreover, we have used single-source self-reported data, which can overestimate energy-saving behaviours. Although this is a widespread approach, some scholars have criticised it. Indeed, self-reported data, especially for behaviours, are vulnerable to social desirability, social approval biases, and retrieval inaccuracy. Future research should include more objective measures, such as tracking of energy use, e.g., via smart plugs (Ruepert et al., 2016). Moreover, although our study considers a prospective measure of behaviour, it is correlational in design and does not allow the inference of causal relations.

In this study, organizational factors or context characteristics were not considered. Future studies could examine which contextual factors affected the likelihood that the TPB constructs or other variables predicted different types of pro-environmental behaviours in workplaces. For example, Lo et al. (2014) found that the number of office co-workers had a negative effect on the behaviour of switching off lights, over and above the TPB constructs. The number of co-workers also had negative effects on attitude and on perceived control over switching off lights. Future research should examine other contextual factors more deeply, e.g., the green-work climate, i.e., employees’ perceptions regarding organisational characteristics and behavioural norms within a company that pertain to environmental sustainability (Norton, Zacher, & Ashkanasy, 2014); perceived organisational support for the environment (Lamm, Tosti-Kharas, & King, 2015); and workplace leaders’ own habits/behaviours and support for green initiatives (Dumitru et al., 2016; Wesselink et al., 2017).

Finally, beliefs associated with attitude towards the behaviours (behavioural beliefs), subjective norms (normative beliefs), and PBC (control beliefs; Ajzen, 1991) were not included. Thus, future studies could examine these beliefs to fully understand behavioural decisions and develop interventions to promote energy-saving behaviours in workplaces.

Nevertheless, our results have certain implications for companies on how they can encourage their employees to adopt energy-saving practices in workplaces. Company management should recognise the role of employees in energy saving and improve their perception of organisational benefits (e.g., showing how much energy and money employees can save for the company if they engage in energy saving). PBC significantly promotes the intention to save energy. Creating a feeling of self-efficacy, i.e., the ability to perform the act, is important in developing energy saving intentions. Thus, interventions to increase PBC are very important. Individuals should be supported to overcome obstacles and barriers by creating favourable circumstances (e.g., providing an energy-saving manual, providing easy access to equipment controls for turning off lights and devices), which should be introduced to develop the intention to perform energy-saving behaviours.

Moreover, low habit to perform energy-saving behaviours in workplaces and low PBC can obstacle the formation of intention and, consequently, the performance of the behaviour. To reduce energy consumption in workplaces, energy-saving behaviours should become habitual in everyday life, which may require the deactivation old habits, particularly for frequent behaviours in relatively stable contexts (Lülfs & Hahn, 2014). In fact, successful habit modification interventions involve breaking up the contextual factors that automatically cue habit execution and introducing changes in the context.