Establishing an energy transition towards higher sustainability is one of the most pressing global challenges. Still, more than 80% of the worldwide energy is produced on fossil fuel basis, leading to CO2
emissions and climate change. Nuclear sources cover another 5%. These sources do not cause CO2
emissions but leave the unanswered question of permanent waste disposal. As compared to these sources, the share of renewable energies in worldwide energy production is rather low (about 14% [1
]). The problems resulting from the current energy production, especially those related to climate change, are visible already, e.g., in terms of extreme weather, floods or changes in ecosystems. If a full energy transition is not initiated immediately, such developments are about to aggravate mostly hitting future generations [2
One crucial element in stopping these developments lies in the adoption of “green” technologies, e.g., highly efficient equipment or systems producing renewable energies; these adoptions imply energy-relevant investments which are known to have a substantial positive impact on the environment [3
]. Understanding how one decides in favor of (or against) an energy-relevant investment will be most helpful for designing policy measures. Unfortunately, there is not much research on energy-relevant investment decisions in social science yet. Only a small number of empirical studies have looked into the determinants of such investments, taking different theoretical perspectives and using various methods [4
Most of the available studies conclude that making energy-relevant investment decisions is more likely if the decision makers expect personal benefits (e.g., cost savings and comfort increases [5
]); these egoistic motives seem to outweigh other determinants—such as eco-social motives or social influences. Recent analyses indicate that these findings may be biased, especially when it comes to social influences. In most studies, social influences are measured via direct questioning and appear to be of minor importance (see [4
] for an overview). However, their decision relevance seems to increase if other measurement approaches (e.g., observations) are used [7
]. The extent of these measurement biases is unclear because there are no studies offering systematic comparisons of different measurement approaches.
In questionnaire studies, it might also make a difference what constructs are used to measure social influences. They can, for instance, either be measured in terms of descriptive (i.e., what significant others are perceived to do [9
]) or injunctive norms (i.e., what others are perceived to expect from the actors’ perspective [9
]). Some studies indicate that descriptive norms have a stronger impact on energy consumption behavior [12
]. However, these studies are rather concerned with everyday behavior and have not yet been replicated for energy-relevant investment decisions.
The scope of this work is to properly capture the role of social influences in energy-relevant investment decisions. We will start with a brief outline of how social influences can be integrated into several well-established action models. We will provide more empirical findings showing that the decision relevance of social influences partly depends on the measurement method. Finally, we will present data from a study in which we compared different measurement methods of social influences. We will focus on PV adoption in households as one example for energy-relevant investment decisions.
4.1. Investment Relevance of Social Influences and Other Investment Determinants
As our first step, we calculated scales for descriptive and injunctive norms, and for the economic, ecological and autarkic motives. We measured all norms and motives with two items; we calculated correlations to check scale reliability. All correlations were statistically significant showing moderate to strong relationships (descriptive norm: r = 0.51; injunctive norm: r = 0.69; economic motives: r = 0.54; ecological motives: r = 0.75; autarkic motives: r = 0.64).
The social norms’ means indicate that descriptive and injunctive norms were both rather unimportant for the PV investment decisions, where descriptive norms were slightly more relevant (descriptive norm: M
= 1.61; injunctive norm: M
= 1.25). All other three investment motives were found to be more relevant. The autarkic motive was rated as most important (M
= 3.58), closely followed by economic motives (M
= 3.45) and ecological motives (M
= 3.44). All descriptive data are shown in Table 1
, which also contains descriptive statistics for each item.
We conducted mean comparisons between the two norms and the other motives in order to gain a better understanding of the differences between the investment determinants. Altogether, 10 mean comparisons were made using paired t
-tests. The alpha level was adjusted to α = 0.005. The results are shown in Table 3
Mean comparisons confirm that descriptive and injunctive norms were less relevant for the PV investment decision than other motives. All mean differences between the norms and the other motives were statistically significant at the adjusted alpha level. Significant mean differences were also found between the two norm types, confirming the higher relevance of descriptive norms as compared to injunctive norms (t(119) = 5.67; p = 0.000). No statistically significant mean differences were found between economic, ecological and autarkic motives.
4.2. Relationships between Survey and Observation Data on Social Influences
We analyzed relationships between the survey data on descriptive and injunctive norms, and observational data (PV systems per capita) by calculating Pearson’s correlational coefficient r. Again, separate analyses were run for descriptive and injunctive norms.
We found no statistically significant correlations between survey and observational data. As compared to each other, the relationship between descriptive norms and observational data (r = 0.06; p = 0.49) was slightly stronger than the relationship between injunctive norms and observational data (r = −0.16; p = 0.86).
5.1. Hypotheses Validation
Hypothesis 1a has been confirmed. Descriptive norms were more relevant for the PV adoption than injunctive norms; the difference between both determinants was statistically significant. Hypothesis 1b needs to be rejected. We only expected injunctive norms—but not descriptive norms—to be significantly less relevant than economic, ecological and autarkic motives; however, we found both injunctive and descriptive norms to be significantly less relevant.
In line with previous research, we found autarkic motives being the most influential determinant for energy-relevant investment decisions [4
]. Autarkic motives could be interpreted as a counterpart to social influences. Especially in individualistic countries, independence is regarded as desirable while being influenced by others is not [23
]. Consequently, the relevance of autarkic motives might be biased by social desirability, if they are measured in self-reports.
Hypotheses 2a and 2b need to be rejected as well, as we found no significant correlations between either descriptive norms and observation data, or injunctive norms and observation data. Especially for descriptive norms, we find these results most surprising. As descriptive norms should reflect the circumstances in the participants’ living area, they should be strongly associated with the observation data. Naturally, self-reported descriptive norms can be biased to some degree [43
], but we would not have expected a zero-correlation.
A major shortcoming of our analyses is that our sample consisted of adopters only. The design would have been stronger if the results had been compared to a non-adopters sample. Such comparisons were not possible in our study: We gathered our data from German PV web portals (see Methods). These portals are used all over the country—and almost exclusively by PV adopters. Thus, it was not feasible gathering a non-adopters group in the same way. Unfortunately, there was no other possibility to gather a comparable non-adopter sample because it would have needed to involve persons from several areas with varying PV diffusion. Future studies should consider designs involving comparisons of adopters and non-adopters but it might be useful to concentrate on certain limited areas in this case.
The sample we investigated was not representative, especially in terms of gender, income, and education of the German population. However, the sample structure is in line with the innovators and early adopters described in the DOI theory [21
] and other PV samples [41
]. These findings are not surprising as these groups are the first to adopt innovations. It might be worthwhile doing more research, probably with other innovative technologies that have already spread further through society. Samples that are more representative could be gathered allowing the comparison between the groups defined in the DOI. It is however unlikely that those social influences are less relevant in these groups. The DOI suggests that late adopters tend to be influenced by early adopting groups who are more likely to be opinion leaders [21
] and can serve as role models for later adopters [44
]. Some evidence pointing in this direction was already found for PV adopters’ everyday consumption behavior [45
]. Indeed, subjective norms were found to predict everyday energy consumption in PV households that purchased their system at a later stage of diffusion.
Additionally, there is room for improvement in the measurements we used. For the survey measurements of social influences and the other motives, we used only two items. That limited number was inevitable as the measurement was part of a larger survey with limited space. In future research, a higher number of items could be used for calculating the scales. In the observations, we only considered PV systems in the participants’ living area but not solar thermal systems as only data for PV were available. Although both systems look different (e.g., different panels and most PV systems are considerably bigger than solar thermal systems), some nonprofessionals may have trouble distinguishing them. For future research in this area, it may be useful considering solar thermal systems as well to see if they also contribute to social influences fostering PV investments.
It should also be noted that the observation of zip code areas may involve some shortcomings. For one thing, different areas may not be entirely comparable. Some might be better suited for PV (e.g., in terms of sunshine hours per day), some may differ in social demographic issues (e.g., income or number of owner-occupied houses), and some might have experienced stronger PV marketing. Such factors may influence PV adoption and should be considered in further studies. It could also be questioned whether zip code areas provide a sufficient resolution to measure social influences. If the area was too large, decision-makers could have trouble to overview the number of PV systems within, and they might also be influenced by systems in bordering zip code areas. Another, probably higher resolution level providing more detailed information (e.g., streets or quarters) would have been desirable—but such data are hardly accessible. Even if they could be gathered, processing them would require extreme effort and involve privacy problems. Additionally, resolution issues might not be too grave at all. For one thing, zip code areas in our analysis were not extremely large. On average, they covered eight square kilometers. Additionally, zip code areas have proven a suitable indicator of social influences beforehand [7
Our research suggests that social influences play some role in energy-relevant investment decisions. Their specific relevance is unclear as it strongly depends on the measurement approach. This finding has several implications mostly from a scientific but also from a practical viewpoint.
Several researchers claim that multiple or mixed methods should be used as one important step to improve research quality in social sciences. (see [46
] for a recent overview). Research focusing on social influences is one area where such approaches are most advisable. Researchers working in this field should be well aware of the constructs (e.g., injunctive vs. descriptive norms), and methods (e.g., survey vs. observation data) that might lead to different results. Studies should always involve several measurement approaches that are compared to each other. Survey data should always cover various constructs. Such measurements should involve descriptive and injunctive norms but also, for instance, interactions with others (e.g., recommendations) prior to the investment decisions. Such determinants could also be measured in qualitative approaches (e.g., [29
]) or experimental designs (e.g., [34
]). Secondary (e.g., observation) data should be also be gathered if possible. In some cases, such data acquisition may cost some effort, but it might be the only way to get a clearer picture of the decision-relevance of social influences.
In our view, the next step is to better integrate different data sources on energy-relevant investments—namely survey and observation data. Most surprisingly, we found survey data on descriptive norms and observation data on PV installations to be uncorrelated. This finding should be investigated in further analyses. Larger samples and comparisons between adopters and non-adopters would be helpful in order to draw better conclusions.
Further studies should also look into other energy-relevant investment decisions and how they are affected by social influences. Similar relationships can be expected for those investments which are as or even more visible than PV investments, such as e-car purchases. Other energy-relevant investment, such as in-home insulations, happen more in private though. Here, social influences might be somewhat less relevant than for the visible investments.
There is also more research needed focusing on other factors than social influences. As we stated before, most of the available studies on energy-relevant behavior are concerned with rather low-impact everyday consumption—and most action models have mostly been verified for these kinds of behavior. We actually cannot really know whether these models also apply for (high-impact) energy-relevant investments but comparative studies indicate that there are some differences [5
]. Rather explorative approaches might be helpful to investigate which factors are relevant and how they interact in order to build a comprehensive conceptual framework explaining energy-relevant investment decisions.
The obvious practical implication of our research is that social influences should be better integrated into policy measures fostering pro-environmental energy-relevant investments. Most policy measures in this area focus on financial aspects where, for instance, funding and low-rate credits are provided [47
]. In addition, some countries also offer professional energy consulting. Energy consulting might always involve some social influences as it generally goes along with personal contact and recommendations. Some analyses already verify the positive effects of energy consulting [34
]; expanding activities in this area might be a promising approach to foster energy-relevant investments. However, energy consultants might not be the best source of social influences as they are usually strangers to the decision makers. Persons who are more familiar and/or in a similar situation (e.g., friends or neighbors) might be more suitable sources of influence. The use of a block leader would be worth considering (e.g., [27
]): persons who already made an energy-relevant investment could be asked to demonstrate it to other interested persons from their neighborhood—probably against compensation. This way, any energy-relevant investment could be presented, and social influences might emerge for those that are less visible than PV use. Such an approach may cost some effort, but social influences and positive investment effects are most likely to occur. Examples supporting this idea can be found in some recent studies: Wolske et al. [30
] found that the interest in talking to an installer was predicted by perceived social support and curiosity about others’ PV systems; Jager [48
] found a positive effect of social networks.
Finally, longitudinal studies might help make policy measures more effective. It should be investigated whether the importance of social influences and other investment determinants truly change by the time an innovation spreads through society—as theory suggests. If so, it might be promising to adopt policy measures’ foci over time.