The effect of social influence in information channels on homeowners’ perceptions of energy efficiency measures

(1)

The effect of social influence in

information channels on homeowners’

perceptions of energy efficiency

measures

Name: Didier J. van Amerongen Student Number: 11956143

Date of submission: 19th June 2018 Submission version: 1.0

Institution: University of Amsterdam, Amsterdam Business School

Qualification: MSc. Business Administration - Entrepreneurship & Innovation First Supervisor: Dr. J. Sol

(2)

Statement of originality

This document is written by Student Didier J. van Amerongen who

declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original

and that no sources other than those mentioned in the text and its

references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the

supervision of completion of the work, not for the contents

.

(3)

1.

2. Abstract

Despite the wide availability of energy efficiency measures for the residential sector, the uptake of these technologies in the Netherlands remains lower than expected. With this study, it is attempted to narrow this so-called energy efficiency gap, which threatens national and global goals to cut carbon emissions. It is analysed how social influence in communication channels affects the homeowners’ perceptions of innovation characteristics and intention to adopt several types of energy efficiency measures. Rogers’ (2003) diffusion of innovations theory was used to study the technologies based on the five innovation characteristics, relative advantage, compatibility, complexity, trialability and observability. An experimental survey was distributed and resulted in 120 elements that were used for statistical analysis. The findings indicate that the intention to adopt energy efficiency measures is related to the five innovation characteristics and that social information provision to some extent is influencing homeowners’ perceptions of characteristics and attitudes towards adoption. This study therefore contributes to current research by showing that social influence is an under-utilized tool for stimulating uptake. At the same time it opens the way for future research to investigate the topic more comprehensively to also obtain more significant results. Additionally, the results suggest that governments and marketers could use social information provision in combination with non-social information at different stages of diffusion to stimulate adoption.

(5)

3. Introduction

The ambition of the Dutch government is to disconnect all dwellings from the natural gas grid before the end of 2050. A daunting task, since it will require over 4.2 million private homeowners in the Netherlands to perform a series of costly energy efficiency measures (EEM’s). In May 2017, the results of a poll held under members of the ‘Vereniging Eigen Huis’ (organisation for homeowners in the Netherlands) suggested that only a low number of homeowners have knowledge about how to perform measures to disconnect from gas (Mulder, 2017). In order to meet the goals set by the European Union, disconnecting from the gas grid and a significant improvement of the energy efficiency of (Dutch) dwellings is needed. Approximately 40% of the total energy and 30% of the total CO2 is used and emitted by buildings in the EU. 30% of these building are residential and consequently they constitute to a large part of the total energy consumption (Filippidou, 2016).

This enormous amount of emissions by dwellings also implies that there is a great opportunity to drastically decrease our carbon footprint by making them more energy efficient. Increasing the energy efficiency of buildings can be done in numerous ways, such as improving weatherization (insulation), installing renewable energy heating systems or installing photovoltaic (PV) panels. It is widely accepted to argue that this opportunity is still underutilized (Häckel et al., 2017). The potential energy savings that can be created and the actual (disappointing) progress that has been made up till now has been referred to in literature as the ‘energy efficiency gap’. The gap can potentially be narrowed by wider adoption of EEM’s by homeowners’ investments, which often can be financially attractive and can have numerous

(6)

co-are suited for improvements, homeowners co-are still not investing in EEM’s at the expected rate (Filippidou, 2016). The gap is not completely understood yet: Studies have focussed on different drivers, barriers, incentives and constraints that explain the complexity of the homeowner’s investment decision, giving contrasting results. Yet, few definitive answers have been found that can really help stimulating EEM investments by homeowners.

According to Owen & Mitchell (2015) EEM’s can be framed as technological innovations to gain a better understanding of EEM investments. Since technologies for closing the energy efficiency gap are already available, and higher energy efficiency is thus possible, faster adoption of the sustainable innovations is the logical and necessary next step for better diffusion. Thus, wider adoption of EEM’s requires an acceleration of innovation diffusion. Owen & Mitchell (2015) discover in their analysis that in the context of innovation diffusion the unexpected importance of intermediaries, such as the role of energy experts, and McMichael & Shipworth’s (2013) findings reveal the importance of innovation diffusion in the context of social networks, or ‘the personal communication network’. The results indicated that information seeking from the social network increased the likelihood of adoption across different types of innovations. Seeking information from a social network will facilitate the adoption of an innovation and thus increase investments in EEM’s. Information that is provided by intermediaries/middlemen and the information that is publicly available will contribute to the knowledge base of homeowners. Since this created knowledge does not always create the desired motivation for EEM investment, this study will instead try to find answers about how the energy efficiency gap can be narrowed by means of information provision through social communication channels. Individuals are educating themselves better on becoming

(7)

more energy efficient. Acquiring truthful and useful information might stimulate the homeowner accordingly to take action. Perhaps, insights into these mechanisms will tell us what information provision approach will be most useful.

Diffusion of innovation theory (Rogers, 2003) will be applied as a theoretical framework to understand the social networks’ role in decision making of households in the adoption of EEM’s. Ultimately, the goal of this study is to increase the uptake of EEM’s of dwellings in the Netherlands. This study attempts to accomplish this by creating understanding about how homeowners’ perceptions of different EEM’s that are currently installable for dwellings in the Netherlands are affecting their intention to adopt them correspondingly. Next to understanding the homeowners and their perception of characteristics, the objective is to find out how provided information through different channels about these measures is influencing their perceptions. The results of this study might give insights in how we can increase the rate of adoption of energy saving technologies among homeowners by selecting the appropriate information channels for the wide set of EEM’s that currently exist. Furthermore, the control variables might provide insights in how socio-demographical differences correlate to adoption intention.

Considering the current state in literature, the highly challenging environmental issue the Dutch housing stock is facing and the goal of this study to understand how homeowners’ perception of energy efficiency measures are related to their intentions to adopt these measures and how these intentions are receptive to social influence, the following research question is developed:

(8)

“How can homeowners in the Netherlands be provided with information effectively to stimulate their intention to adopt energy efficiency measures?”

By answering this research question, several theoretical contributions will be made. The results will:

 Contribute to the ‘energy efficiency gap’ theory.

 Contribute to the ‘Diffusion of Innovation’ theory and creating a further understanding of consumer perceptions in the context of energy efficiency measures.

 Contribute to theories of social information provision.

Besides contributing to theory, this studies’ research question will provide insights for the sector about how the homeowners should be informed and could be convinced in investing in their dwellings in order to increase the energy efficiency. Insights could help designing energy efficiency interventions more effectively to speed up the transition to energy efficient dwellings. The contribution is therefore relevant for policymakers and governments, as well as for businesses operating in the sector, since they will gain insight in how their marketing strategies should be designed. Furthermore, the results will also be of interest to homeowners, since these provide insights in how relevant information about EEM’s can be transferred to them more efficiently. Accordingly, the result will be particularly relevant for the residential sector in the Netherlands.

The first part of this study will elaborate and discuss prior literature on this topic, to develop a better understanding thereof. By determining the background for this study,

(9)

the foundation for the research question, conceptual model and hypotheses will be laid and subsequently will be presented. After this clear delineation, the second part of this study will describe the methodology of this empirical research. The design of the survey, a description of the measures used and a description about the data collection will be included. This study then proceeds with a presentation of the results following from data analysis and will discuss what these results entail. Finally, the conclusion will complete the study.

(10)

4. Literature review

This chapter provides an overview of the current literature in the field of energy efficiency in the residential sector and a review of the theory of innovation diffusion.

4.1 Innovation Diffusion

A much referred to theory about diffusion of innovations is the work of Everett Rogers. Innovation diffusion theory attempts to explain how social systems contribute to the spreading and adoption of ideas and new technologies. He defines it as: “the process in which an innovation is communicated through certain channels over time among the members of a social system” (Rogers, 2003, p. 5). According to Rogers’ definition, innovation diffusion thus consists of the 4 ‘essential elements’: The innovation characteristics, the communication channels, (adoption over) time and the social system. They form the basic dimensions in which innovation diffusion can be studied.

An individual that adopts an innovation follows a five-step process (Knowledge, Persuasion, Decision, Implementation, and Confirmation). During these five stages an individual can reject or accept the innovation. The five-step innovation decision process starts with the knowledge stage, where the individual learns about the existence of the technology and seeks further information. Persuasion, the second stage, involves the consumers’ evaluation of the relative advantages, price and complexity of the innovation. The individual forms an attitude towards the innovation, which is more emotional, and subjective in nature and thus, the favourable or unfavourable attitude towards the innovation is strongly influenced by peers whose subjective opinions are most convincing. The individual’s assessment in this stage

(11)

results in a high or low intention to adopt the innovation. After this evaluation of the innovation, the individual will decide in the third stage whether to make use of the innovation or not. After the decision stage, the individual will in the implementation stage actually purchase and use the innovation. At the final confirmation stage the innovation decision has been made, but the individual seeks for reinforcement of his or her decision.

4.2 The relevant diffusion stages for this study

The EEM-information can be provided to homeowners through multiple communication channels in either the knowledge or persuasion stage (stages 1 and 2) and the homeowner might still reject or postpone the adoption. Obviously, for the rate of adoption to increase, the homeowner must accept the innovation in the decision stage (stage 3). Though, indecision under homeowners does not mean that the rate of adoption decreases, since the homeowner might seek for further information, or delay the decision and come to the decision of acceptance at a later period in time. To understand the effectiveness of information provision on the homeowner’s intention to adopt an EEM, this study will consequently look at how the information provided in the knowledge and persuasion stage results in the homeowner to stay motivated enough to proceed with the adoption process. Which is, seeking for further information (stage 2) or weighing the advantages and disadvantages of the particular EEM to eventually come to a decision (stage 3).

4.3 Characteristics of the innovation

(12)

rate of adoption. The individual learns about the innovation’s characteristics through communication channels. The 5 most important characteristics are;

1. Perceived relative advantage: The perceived degree to which the innovation is more advantageous than the one it displaces. Relative advantage can cover a large amount of benefits, such as social benefits, economical benefits, convenience, product quality and environmental benefits (Flight et al., 2011). In the context of energy efficiency, a relative advantage is also the independence from the electricity and/or gas grid. A Dutch study revealed that 40% of the households have an intention to generate their own power, and thus, this attribute can be seen as an important relative advantage for homeowners (Leenheer, de Nooij & Sheikh, 2011).

2. Perceived compatibility: The perceived degree to which the innovation is consistent with current needs, values and norms of the adopter. Incompatibility requires the adoption of new values, which is a slow process and thus, will cause slower adoption of the subsequent innovation.

3. Perceived trialability: The degree to which the innovation can be tried out or be experimented with. Trying out products before purchase decreases uncertainty for adoption.

4. Perceived observability: The degree to which results of the innovation are visible. The visibility of the results will stimulate discussion among peers. Solar adopters are for instance often regionally clustered, because adopters can more easily request innovation-evaluation information about the product from their neighbours or friends.

5. (Less) perceived complexity: The perceived degree to which the innovation is difficult to understand and use. The more the innovation requires the adopter

(13)

to develop new skills or understandings, the less rapid the adoption will take place. Higher perceived complexity of the innovation, in contrast to other product characteristics, is generally associated to lower motivation and intention to pursue the adoption of the innovation.

Other characteristics can be found in literature, such as environmental friendliness and the risk of uncertain performance (Claudy et al., 2011). Where environmental friendliness of EEM’s can be interpreted as a relative advantage. The majority of innovation diffusion studies adhere to these five categories as good predictors of innovation adoption rates. Of the five categories, relative advantage is generally found to be the best predictor of individual adoption of innovations.

4.4 Focusing on adoption groups

Individuals, or adopters, are classified within different types of groups: innovators, early adopters, early majority, late majority and laggards. All groups possess different characteristics, like personality traits and socioeconomic status (Rogers, 2003). The distinction is made to help understand and predict how innovations are spreading in social groups.

The theory of innovation diffusion can be applied by creating strategies to increase the rate of adoption of technological innovation in order to reach ‘critical mass’. Critical mass refers to the sufficient number of adoptions, or level of diffusion, where the innovation becomes self-sustaining. The marketing strategies that will be created, should focus on the specific subgroups, by keeping the following drivers per subgroup in mind: Innovators are driven by technology enthusiasm, early adopters by

(14)

pressure and laggards only by force (Rogers, 2003). The study conducted by Diaz-Rainey & Ashton (2015) presented strategies for adoption of nine EEM’s, such as double and triple glazing windows and weatherization measures. They found that the adopter profiles are consistent with the characteristics given in innovation diffusion theory. This knowledge can therefore be used for better targeting of these social groups to stimulate the adoption of EEM’s.

4.5 The Context: Energy Efficiency Measures and Social Networks Enhancing energy efficiency has been a topic of research for years. The earliest notion of the influence of a social network on diffusion of EEM’s in the built environment is found in the paper of Darley (1978). It brought to the attention the role of social techniques to guide homeowners towards investing in EEM’s. A lot of work has been performed also more lately on studying the effect of providing information to homeowners to increase consciousness and perception of benefits (Friege & Chapin, 2014; Henryson et al., 2000; Stiess & Dunkelberg, 2013). However, the effect of information provision and educating of homeowners does not deliver reliable results in terms of behaviour change towards EEM’s (Cole et al., 2018). According to Cole et al. (2018), the role of information provided within social networks is much bigger. This is also found in other studies. In the UK for instance, households often gather information through their social networks, which potentially is used in the decision-making (McMichael & Shipworth, 2013). This is also the case in the US, where the social information exchange about energy measures, like weatherization (insulation), are more related to these measures than the actual level of education of a homeowner (Southwell & Murphy, 2014). Kastner & Stern (2015) also underpin the importance of social sources, which are probably more trusted, and recommend

(15)

further research to find out what the exact influence of this source of information is within the investment decision process. Consequently, these channels can be used to increase the rate of diffusion by providing the relevant information for decision-making.

The conducted studies by McMichael & Shipworth (2013), Southwell & Murphy (2014) and Kastner & Stern (2015) delivered similar recommendations for policy and marketing to improve diffusion. They argue that investing in information dissemination through social networks will be an effective policy, if well targeted to specific communities. After all, information exchange is associated with adoption and therefore should be triggered. Additionally, Bale et al. (2013) simulate a network model (in the UK) and find proof that network effects enhance EEM uptake and suggest that encouraging the communication of information in networks can therefore lead to higher uptake. Their results further suggest that early adopters are ‘needed’ for this spread of information by influencing others.

Another study conducted by Bollinger & Gillingham (2012) illustrates how social interaction (in California) leads to PV panel adoption. Visibility of the panels (observability), and word-of-mouth lead to an increase of interaction between peers, which in turn lead to higher adoption rates. The paper shows that for the PV market, and possibly other visible green technologies, interaction can be stimulated among social groups by providing the correct, and positive information (i.e. marketing). The effect is present, because for EEM’s adopters gather information through their peers’ experiences, since the technology cannot be tried out first-hand before purchasing (Rogers, 2003). Low trialability of EEM’s is a partial explanation for their slow uptake.

(16)

When marketers intervene in social networks, they should understand that social influence could have a positive as well as a negative effect on the diffusion of the innovation, especially when the critical mass is not going to be reached. In more uncertain markets, with high social influence, the focus should be on small periods and regions of the market to convert this social influence from a negative to a positive one (Delre et al, 2010).

4.6 Information provision and communication channels

The effect of information that is provided on ones motivation to adopt EEM’s has been shown in studies. Aravena et al. (2016) find that homeowners that are more satisfied with the information provision are also 5,5% more likely to adopt EEM’s, compared to unsatisfied homeowners. Therefore, information provided should be accurate and appropriate.

Pro-environmental behaviour, need for saving money, enthusiasm for the technology and co-benefits, like increased comfort, are just a few of the motives for homeowners to adopt EEM’s and the relative importance of arguments differ individually and demographically (Cole et al, 2018; Pelenur & Cruickshank 2014). Homeowners generally are not sufficiently aware of all the benefits of EEM’s, due to a lack of information provided to them and mismatched information campaigns (Aravena et al., 2016). Although Aravena et al. (2016) call for further improvements of how information is designed, the unavailable information and ineffectiveness of the information might be due to the communication channel through which the information is provided (Rogers, 2003). According to Rogers (2003), individuals are especially seeking for information to evaluate the innovation in the persuasion stage via their near-peers, because this lowers the uncertainty about expected consequences

(17)

of the innovation’s adoption. The channel through which the homeowner is informed about EEM’s might accordingly influence the motivation to adopt. This has also been found in different studies (Bollinger & Gillingham 2012; Delre et al., 2010; McMichael & Shipworth, 2013). Information about the economic benefits of an EEM might increase one’s motivation to adopt, but the effect might be stronger when such information is provided through a social channel, since it increases information satisfaction. This is because the information is more likely to be perceived as useful (Palm, 2017) and trustworthy (Wolske et al., 2017) it can increase motivation to adopt certain types of EEM’s. This increases the individual’s perception of benefits of the EEM, which will improve the rate of adoption.

4.7 Intention to adopt

Innovation diffusion theory explains the spread of innovative ideas among the members of social systems, but does not provide a satisfactory method to measure the effect of an intervention on innovation adoption in individuals. To measure the motivation of individuals to adopt innovations, a commonly used substitute, as seen in earlier literature, is intention to purchase (Holak, 1988; Korcaj et al., 2015). Insights in intention to purchase are valuable because a higher intention increases the chances of actual purchasing behaviour (Ajzen, 1991). Intention, as a strong determinant of behaviour, provides valuable insights about the individual. However, it is not in this study’s interest to further investigate exact characteristics of intention, such as ‘perceived behavioural control’, that are present in individuals. These attributes are widely discussed in Ajzen’s theory of planned behaviour, which to some extent overlaps with Roger’s Innovation diffusion theory (Wolske et al., 2017), and are used

(18)

in predicting intention. Instead, the goal of this study is not to predict the inflicted effect in individuals’ interest in EEM, but to measure it.

4.8 Classifying energy efficiency measures

A large number of EEM’s exist and are being installed in the Netherlands. Testing all the different EEM’s in this study would require a large sample and moreover would be inefficient. Many of these measures hold similar product characteristics in the perception of the homeowner. Labelling the EEM’s as innovations therefore enables the researcher to test a selected amount of measures that cover all the characteristics appropriately. As a result, only a selected amount of measures have to be tested, instead of the full set of measures that exist. Table 1 shows for each of the measures the scores per characteristic. The scoring of the characteristics was done by collecting information found on informative websites (f.i. www.milieucentraal.nl), and where possible, by analysing previous studies. For instance, the scores for PV panels where adopted from Vasseur & Kemp (2015) and the scores for the solar boiler and wood/pellet-boiler were derived from the study of Claudy et al. (2011) The scores in table 1 range from highly negative “--“, to highly positive “++”. Based on the analysis of the different EEM’s, the researcher selects 4 diverse measures that cover the range of product characteristics. The 4 selected measures are: floor insulation, air-source heat-pumps, double glazing (windows) and solar boilers. The selected measures are highlighted in table 1. The table also presents the mean scores (on a 7 point Likert scale) of the homeowners’ perceptions of innovation characteristics that were subsequently collected from the empirical research.

(19)

Note: The numerical values in bold represent final scores from the empirical research.

4.9 Conceptual Model

The conceptual model is presented in figure 1. The scope of this study ends at the ‘Homeowner’s intention to adopt EEM’. I found a reasonable amount of studies that support that an intention to adopt EEM will eventually lead to diffusion of the EEM. It will not be possible to perform a longitudinal study for this thesis due to time

Table 1: Evaluation of Energy Efficiency Measures

Energy Efficiency Measure

Relative advantage

Observability Complexity Trialability Compatibility

Air-source heat-pumps ++ (5,1) - (4,8) ++ (4,4) -- (4,4) - (4,6) Cavity wall insulation + - +- - + Double glazing (windows) +- (4,4) + (5,4) - (2,6) + (4,8) + (5,6) Floor insulation +- (4,3) +- (5,4) - (2,6) +- (4,7) + (5,5) Ground-source heat pump ++ - ++ -- - Hybrid heat-pumps + + + - - Infra-red panels - + +- + - LED-lights - ++ -- + - PV panels + ++ - - + Roof insulation + - + -- + Solar boilers + (4,8) ++ (5,0) +- (3,2) - (4,2) + (5,4) Underfloor heating +- + +- + + Wood/pellet-boilers + +- - + +-

(20)

EEM’s. The purpose of this study is to measure the effect of information provision on the diffusion process at the individual level. Therefore, it is also possible to measure the individual’s motivation to proceed with the adoption, or the intention to adopt. After all, as was reasoned in the literature review, once an individual intends to adopt the innovation, it is more likely that the individual will eventually purchase the innovation concerned. While it would be interesting from a practical point of view to obtain measures of actual adoption, an increased intention to adopt is also a satisfactory measure of innovation characteristic’s influence in the diffusion process. Claudy et al. (2011), for the same reason, used this approach to measure the influence of innovation characteristics on ‘Willingness to Pay’ (WTP) in the specific context of micro-generation technologies to gain valuable insights for policymakers and marketers to promote the uptake of this technology. Though, WTP will not exist as the dependent variable in this study, since WTP represents actual consumer behaviour following from the purchase intention (Luzar & Cosse, 1998; Schniederjans & Starkey, 2014). WTP measures the (numerical) amount an individual would like to pay for a product, which exceeds the intended goal of measuring the increased chance of purchase. For this reason, this study will investigate the dependent variable ‘intention to adopt. Intention to purchase, or intention to adopt is considered to be a relevant measure for the expected effect resulting from manipulation of the information channel on perception of innovation characteristics.

Undoubtedly, there are several barriers and thresholds that will lower the rate of diffusion, such as constraining features of the home itself or an individual’s lack of purchasing power etc. Furthermore, other drivers for ‘Homeowner’s intention to purchase EEM’ are also not being studied in further depth. For instance, personal beliefs and preferences, like ‘environmental motivation’, are not being studied. Due to

(21)

time constraints this is being left out of scope. Thus, the study will focus on the intention to purchase, the perception of characteristics and the moderating effect of the homeowner’s information channel between these two variables.

Figure 1: conceptual model

Hypotheses:

For the hypotheses, the ‘perception of characteristics’ consists of the five parts relative advantage, compatibility, complexity, trialability and observability. Four out of five are positively related to the dependent variable. Only complexity has an expected opposite (negative) effect.

H1a: There is a positive relation between a homeowner’s positive perceptions of

(22)

H1b: There is a negative relation between a homeowner’s positive perceptions of

EEM’s complexity on the homeowner’s intention to adopt energy efficiency measures.

H2a: The positive relation between a homeowner’s positive perceptions of EEM’s

relative advantage, compatibility, trialability and observability on the homeowner’s intention to adopt energy efficiency measures is moderated by the social level of the information channel, so that this relationship is stronger for higher socialness of the channel.

H2b: The negative relation between a homeowner’s positive perceptions of EEM’s

complexity on the homeowner’s intention to adopt energy efficiency measures is moderated by the social level of the information channel, so that this relationship is weaker for higher socialness of the channel.

H2c: The higher the socialness of the information channel, the more positive the

individual’s perceptions of relative advantage, compatibility, trialability and observability, and, the more negative the perception of complexity.

(23)

5. Methodology

This chapter will explain the methods used in this study to find an answer on the main research question and includes the design of the study, the measures used, details about the sample and raw data.

5.1 Design of study

In order to find an answer on the research question and to correctly measure the hypotheses, the study will attempt to find out how homeowners’ motivations are affected by information that they obtain from information channels. To measure the effect of the information channel, two samples are created. One sample obtains their information through a social channel, the other sample through a non-social channel, framed like an internet-webpage. The two surveys that are created (high/low social level of information provided) will be manipulated is such a way that the effects on homeowner’s motivation can be measured.

The quantitative research will be carried out using a cross-sectional experimental survey design and involves a between subjects design with 2 treatments. This type of research is useful for using statistics to test the two hypotheses. Two different questionnaires will be administered digitally and respondents will be assigned randomly. The survey will ask about the homeowners’ perceptions of innovation characteristics and about their intention to purchase the shown EEM. The survey will also enable the researcher to easily and efficiently collect data about the individual’s demographics.

(24)

The administered questionnaire consists of two parts. Before a respondent could participate, a check was performed to make sure that the respondent belonged to the target group of this study, the Dutch private homeowner. After this control question the survey started with part one, where data about age, gender and education was collected about the respondent to create a personal profile. Part two consists of the two information materials and corresponding questions to measure the individual perception of characteristics and intention to adopt the shown EEM. Respondents were first assigned randomly to the social or non-social group to create the between-subjects condition. Within the social or non-social group, the respondent was then assigned randomly to two of the in total four researched EEM’s to create the within-subjects condition.

The design of the information that is provided within the experimental survey can be found in Appendix A. Information about EEM’s provided in both surveys, as well as the questions that followed, are nearly identical. However, the communication channel of the information is manipulated. The information provided to respondents was selected to be neutral, which is, not in favour or in some way negative about the specific EEM. Participant of study 1 was shown a screenshot of the fictitious Facebook friend Frans, who recently shared information about an EEM. The social information was presented as coming from the person ‘Frans’, which included a profile picture. Participants of study 2, the control group, was shown a screenshot of a fictitious Facebook-page with the name; ‘Duurzaam Wonen NL’ (English: ‘Living Sustainable NL’), which included a picture of energy-labels for dwellings. This page is profiled as a neutral advisory company providing information about subjects like EEM’s. Additionally, the social channel (‘Frans’), included more informal language to create a more personal ambiance towards the respondent. The shared pieces of

(25)

information are kept short, since lengthy questionnaires tend to have different disadvantages. Dropout rates might become higher, attention levels could drop and information might be processed less well (Fan & Yan, 2010). Figures 2 and 3 show for one EEM the non-social (figure 2) and social (figure 3) treatment. The remaining 6 information designs can be found in Appendix A.

Figure 2. Non-social information about the solar boiler

(26)

The different characteristics were carefully tested by deliberately selecting specific information. For all the EEM’s, the extent to which advantages and disadvantages were mentioned was carefully considered, creating a wide spectrum of perception of the relative advantages. An additional example, for the air-source heat pump, information was provided about some complex conditions that need to be present in a homeowner’s dwelling. This is done to represent the characteristic ‘complexity’, which is evaluated to be high for the air-source heat-pump in table 1.

5.2 Measures

Homeowners responded to survey items using a seven-point Likert type scale, ranging from 1 = “strongly disagree” (Dutch: ‘zeer mee oneens’) to 7 = “strongly agree” (Dutch: ‘zeer mee eens’). The survey was administered in Dutch. I used a back-translation method to translate all items. Some of the measures are presented in this chapter. A complete overview of all the items can be found in appendix B.

The independent variables, which are the 5 characteristics of Rogers (2003), were assessed using adapted questions of the measurements in the study of Claudy et al. (2011) and Van Ittersum & Feinberg (2010). The contents written in between the brackets (“< … >”) where adjusted to the words “energy saving measure” (Dutch: ‘energiebesparende maatregel’) and then the full items were translated to Dutch. The measurements included, “<Micro wind turbines> are very complex products” (Cronbach’s α = 0.78) to measure complexity, and “You could draw on someone’s experience who has installed a <small wind turbine> already” (Cronbach’s α = 0.68), to measure trialability. As noted in the theory section, perceived relative advantage can consist of numerous subjects and could therefore also be perceived differently by homeowner. Following the approach of Claudy et al. (2011), the scale ‘relative

(27)

advantage’ contains five elements representing environmental friendliness and independence from the energy grid. While it would also be interesting to measure economical benefits as a substance of relative advantage, choices had to be made. Taking into account that price-related content is generally more tailor-made to a homeowner’s specific situation, providing information thereof and testing this attribute could also decrease the veracity of this study. Furthermore, adding this element would further increase the length of the questionnaire, which is undesirable. One of the elements used was “By installing a <micro wind turbine> on your house you would help to significantly reduce greenhouse gases.” (Cronbach’s α = 0.88). The dependent variable was measured using Brown & Venkatesh’s (2005) scale. An example of an item is; “I intend to adopt a <computer> at home” (Cronbach’s α = 0.90).

To rule out potentially spurious relations, in all the analyses control variables for age (in years), gender (1 = male, 2 = female) and education level (1 = Primary school, to 5 = University level) were assessed, since these control variables are common in similar studies in the field of innovation diffusion (Balcombe, Rigby & Azapagic, 2013; Claudy et al. 2011; Diaz-Rainey & Ashton, 2015; Franceschinis et al., 2017; Prete et al., 2017; Vasseur & Kemp, 2015). It is found for instance that higher educated people in the Netherlands tend to be more eager to adopt innovative products and therefore also have a higher intention to adopt (Vasseur & Kemp. 2015).

5.3 Sample and data collection

The population of interest in this study are all the Dutch private homeowners. This population is large and contains about 4.4 million people in the Netherlands. Since

(28)

convenience sample and a non-probability voluntary sample. 362 respondents were reached through a short and attractive message in a newsletter that was send to the customers in the database of a company that is active within in the energy efficiency sector. Also convenience sampling was part of the data collection. The survey was distributed among the personal network of the researcher by placing a general invite on LinkedIn and Facebook and sending personal invites via Whatsapp and e-mail a week after the general online posts.

The researcher attempted to collect as many respondents as possible during the data collection period. Since the sample must consist of as many respondents as possible and the available resources were limited, the chosen method of survey distribution was a self-administered questionnaire spread via the Internet. Respondents were invited to follow a link to ‘Qualtrics’, a widely renowned software company for collecting data anonymously. To make sure that participants of the survey were actual homeowners and accordingly matched the population of interest, a check-up question was asked right at the start of the survey. Of the 163 who started with the questionnaire, 14 responses where filtered out due to this check.

Combining these two techniques resulted in a total number of respondents of 163 who started the questionnaire. From the 149 residual responses, another 29 responses were deleted, since not all responses were completed and some were invalid. Resulting in a dropout rate of 19.4%. Some incomplete responses could still be used for analysis, since some data was missing completely at random and therefore could be edited for analysis. 21 of the 120 final respondents only answered the questions for one EEM, and skipped the second series of questions. These results from partial completion are included in the data analysis, since half of the results could still be used. Additionally, some cases had larger amounts of missing values in

(29)

only one EEM-variation and could therefore be interpreted as partial completion. After cleaning the data from unusable responses, a total of 219 elements (99*2 + 21) could be used for analysis. Furthermore, few missing values were detected in the dataset. The missing values were assumed as missing completely at random. In order to reduce bias in the results as much as possible, only the cases without missing values were analysed (listwise deletion). From the 120 unique responses that could be used for analysis, the number of responses from the convenience sampling was 116. The number of responses from the newsletter was rather low, namely 4 (rate of response 1.1%). The final sample, eventually, consisted of 120 elements of which 64% were male and the average age was 48 years old

(30)

6. Results

For the analysis of the hypotheses, a dummy variable was created to make a distinction between the respondents who were randomly assigned to the social and non-social information condition. The variable ‘information channel’ contained the values ‘non-social’ (0) and ‘social’ (1). In order to gain insights in the dataset, the mean value was computed for all independent and dependent variables of all the 4 tested EEM in both the social and non-social variation. This resulted in the values presented in table 2, showing both the mean values and standard deviations for all 48 variables.

Table 2: Mean and Standard Deviation of Variables

Variables EEM-Type Relative Advantage Compati-bility Triala-bility Com-plexity Observa-bility Intent to adopt Non-soc. Social Non-soc. Social Non-soc. Social Non-soc. Social Non-soc. Social Non-soc. Social Floor Insulation N 25 29 25 29 25 29 25 29 25 29 25 26 Mean 4,35 4,26 5,63 5,45 4,66 4,69 2,44 2,82 5,64 5,25 3,53 3,60 SD 1,18 1,12 0,81 1,06 1,47 1,31 0,89 1,34 0,74 1,04 1,48 1,68 Heat Pump N 27 29 27 28 27 28 27 28 27 28 26 28 Mean 5,06 5,08 4,84 4,43 4,80 3,93 4,23 4,48 5,12 4,55 3,99 3,75 SD 0,96 0,98 1,56 1,52 1,15 1,41 1,27 1,51 1,21 1,40 1,86 1,70 Window Insulation N 23 31 23 31 23 31 23 31 23 31 23 31 Mean 4,34 4,39 5,42 4,39 4,98 4,68 2,90 2,45 5,38 5,37 3,88 3,90 SD 1,00 0,99 1,12 0,99 1,35 1,29 1,06 1,06 1,25 1,09 1,71 1,45 Solar Boiler N 28 28 28 28 28 28 27 28 27 28 28 28 Mean 4,71 4,84 5,26 5,52 4,34 3,91 3,23 3,11 4,88 5,19 2,99 3,69 SD 0,93 1,19 1,09 1,00 1,60 1,26 1,20 1,22 1,25 0,98 1,72 1,47 Note: Values in bold tested significant on an independent sample t-test (p < ,05)

(31)

Although the small sample size has low statistical power, these results indicate some interesting differences between the perceptions of characteristics between the different selected EEM’s. Additionally, the social manipulation of information provision appears to influence some variables, some more substantial than others. The intention to adopt a solar boiler seems to be affected by social influence to some extent, while the other EEM’s show less variance. These results show that the perceptions of characteristics and intention to adopt an EEM vary across the tested between-subjects condition. This outcome can be shown in a graph to create better understanding of the effect. Figure 2 shows for each EEM, the average intention to adopt on the 7-point Likert scale in both the social and non-social group. We can now easily see that the average intention to adopt EEM’s converges. Furthermore, floor insulation and window insulation seem to show less variation than the solar boiler, which has a notable positive upward line, and the air-sourced heat-pump, which shows a small downward trend towards adoption under social information provision.

2.50 3.00 3.50 4.00 4.50 Non-Social Social Int en tion to Ado pt

Mean Scores, Intention to Adopt

Floor Insulation Heat-Pump Window Insulation Solar Boiler Average

(32)

Reliability of the scales

To test the reliability of the scales, the Cronbach’s alpha scores of the scales are computed. Four of the five perceptions of characteristics scales have high reliabilities, with Relative Advantage, Compatibility, Complexity and Observability having a Cronbach’s Alpha of .73, .76, .81, and .72 respectively. The corrected item-total correlations indicate that these four items have a good correlation with the total score of the scale (> .30). Also, none of the items would substantially affect reliability if they were to be deleted. Trialability has low reliability, having a Cronbach’s Alpha of .56. Since the scale only consists of two items, deletion of one of the items cannot be performed to increase reliability. The used scale for the dependent variable, “intention to adopt”, also has a high reliability of .91. The corrected item-total correlation scores are sufficient.

Validity of the scales

To test validity of the scales used for the independent variables, a principal axis factoring analysis (PAF) was conducted. The sampling adequacy for the analysis was successfully verified with the Kaiser-Meyer-Olkin measure (KMO = .761). The scores of Bartlett’s test of sphericity χ² (120) = 1111.219, p < .001, reveal that there were also sufficiently large correlations between the items for PAF. Furthermore, analyses were run to obtain the eigenvalues for the components. A total of 5 components had eigenvalues that passed the Kaisers criterion, which altogether explained 65,45% of the variance. Although, the scree plot revealed a levelling off after the third factor, the fourth- and fifth factor were suitable for analysis in an Oblimin rotation with Kaiser normalization. Resulting in 5 factors representing the 5 perceptions of innovation characteristics. The results of this analysis are shown in

(33)

table 3. Clustered items on the same factor suggest that factor 1 represents perceived compatibility, factor 2 perceived relative advantage, factor 3 perceived trialability, factor 4 perceived complexity and factor 5 perceived observability.

The results also suggest some higher cross-loadings for Trialability with the items REL1, REL2 and OBS16. This corresponds to the results from the reliability test, which already revealed the low strength of the scale. The overlap is possibly explained by the fact that the degree to which an innovation is ‘trialable’ (or easy to see/try first-hand) is also related to the degree to which an individual clearly understands the benefits of the EEM, and also, the content of REL1 and REL2 being particularly focussed on environmental advantages which are possibly relatively easy to perceive by homeowners. Taking a closer look at the related item OBS16, which asks if the results of the innovation are apparent to the homeowner, we see that the main difference from the other two observability items is that it does not include communication with others about the results. This suggests that the homeowners understand the EEM’s results within their own situation, but not for that of others. For instance, it might be difficult to explain outcomes for innovative products that cannot be tried out first hand and that have uncertain or unknown environmental advantages. Another cross loading was found on the item CLX12. This item asked homeowners if they thought the EEM would be difficult to use, but does not load only on complexity, but additionally had a factor loading of -.31 on compatibility. A possible explanation for this factor loading is the fact that the item can be linked to the homeowner’s capability to integrate a complex product into ones lifestyle.

(34)

Table 3: Factor Loadings Items**

Rotated Factor Loading* Compatibility Relative

Advantage Trialability Complexity Observability

REL1 0,35 0,48 REL2 0,31 0,55 REL3 0,81 REL4 0,87 REL5 0,75 CPB6 0,75 CPB7 0,81 CPB8 0,76 TRL9 0,76 TRL10 0,61 CLX11 0,84 CLX12 -0,31 0,66 CLX13 0,87 OBS14 0,83 OBS15 0,84 OBS16 0,47 0,42 Eigenvalues 4,01 2,64 1,67 1,10 1,05 % of variance 25,08 16,52 10,41 6,85 6,58

*Note: Factor loadings < .30 were suppressed **Note: For the content of the items, see Appendix B

(35)

Analysing the intention to adopt

Since this study primarily aims to find out how the homeowners’ perception of characteristics predict the intention to adopt, a correlations matrix and a regression analysis were carried out. To start of with, the correlations between the dependent and independent variables, control variables and moderator are shown in Table 4. There was a small to medium positive correlation between the four independent variables relative advantage (r = .26, p ≤ 0.01), compatibility (r = .29, p ≤ 0.01), trialability (r = .30, p ≤ 0.01), and observability (r = .21, p ≤ 0.01), and the dependent variable intention to adopt. The variable complexity had a small negative correlation with the variable intention to adopt, r = -.19, p ≤ 0.01. Furthermore, a small positive correlation was found between the two variables age and intention to adopt. It should also be noted from these results that correlations were found between the control variables. These correlations point out that the sample in this study, as expected, is somewhat biased and contains a disproportionate amount of higher educated older males. As we can also read, being a female correlates to higher intention to adopt. Also, higher age has a small correlation with perceived trialability, and higher educated homeowners correlate negatively with perceived relative advantage, compatibility and trialability. Lastly, the socialness of the information channel (control = 0, social treatment = 1) has a weak and negative correlation with trialability, which is an unexpected result.

(36)

Table 4: Correlations Variables (N = 219) μ SD 1 2 3 4 5 6 7 8 9 10 1. Gender (M = 1, F = 2) 1,36 0,49 2. Age 48,2 9,93 ,14* 3. Education level 4,32 0,86 -,30** -,23** 4. Relative Advantage 4,64 1,08 0,09 -0,06 -,14* (.73) 5. Compatibility 5,29 1,25 0,11 0,05 -,15* ,17* (.76) 6. Trialability 4,49 1,39 0,05 ,20** -,18** 0,11 0,12 (.56) 7. Complexity 3,21 1,40 -0,01 0,07 -0,03 0,11 -,45** -,14* (.81) 8. Observability 5,17 1,16 -0,09 0,02 -0,03 ,163* ,35** ,37** -,34** (.72) 9. Intention to Adopt 3,66 1,64 ,18** 0,04 -0,13 ,26** ,29** ,30** -,19** ,21** (.91) 10. Information Channel 0,53 0,50 -0,04 -0,08 -0,04 0,01 0,01 -,13* -0,01 -0,06 0,05 * Correlation is significant at the 0.05 level (2-tailed).

(37)

To further predict the ‘intention to adopt’ under homeowners, a hierarchical regression analysis was run. The saved factor-scores from the factor analysis are used in this analysis to represent the five perceived innovation characteristics. The results of the regression are shown in table 5. Firstly, in this hierarchical multiple regression there was controlled for age, gender and education level. The first model included the demographic variables and was found to be statistically significant; p < .05, F (3,208) = 3,01. 4.2% of the variance in intention to adopt was explained by the control variables. The second model was then tested and also included the five independent variables relative advantage, compatibility, complexity, trialability and observability. This model was also found to be significant, p < .001, F (8,203) = 6,21. The second model explained 19.7% of the variance in intention to adopt. Thus, the five innovation characteristics in total added 15.5% (R2 change = .155, F (5,203) = 7,83) of the variance, after controlling for age, gender and education level. Within the second model, four out of eight variables were statistically significant. From the control variables, only gender tested significantly (β = .14, p < .05). Three out of five independent variables tested significantly, namely relative advantage (β = .15, p < .05), compatibility (β = .17, p < .05) and trialability (β = .20, p < .01). These results can be interpreted as follows; if perceived relative advantage is increased with one unit, the intention to adopt increases with 0.24 units and 0.15 standard deviations (SD), one unit of compatibility with 0.28 and 0.17 SD, one unit of trialability with 0.33 and 0.20 SD. Furthermore, observation of the results reveals that women on average obtain a 0.55 units and 0.16 SD higher score on the intention to adopt than men. The third model included also the socialness of the information channel. This model did not test significantly (p = ,24 > ,05). This result suggests that the increase

(38)

in the socialness of the information channel does not explain variance in intention to adopt.

Table 5: Hierarchical Regression Model of Intention to Adopt.

R R^2 R2 Change B SE β t Model 1 0,20 0,04 0,04* Gender 0,55 0,24 0,16* 2,29 Age 0,00 0,01 -0,01 -0,15 Education Level -0,17 0,14 -0,09 -1,21 Model 2 0,44 0,20 0,16*** Gender 0,47 0,23 0,14* 2,03 Age 0,00 0,01 0,01 0,08 Education Level -0,04 0,13 -0,02 -0,27 Compatibility 0,28 0,11 0,17* 2,43 Relative advantage 0,24 0,11 0,15* 2,21 Trialability 0,33 0,11 0,20** 2,92 Complexity -0,18 0,11 -0,11 -1,62 Observability 0,15 0,11 0,09 1,30 Model 3 0,45 0,2 0,01 Gender 0,48 0,23 0,14* 2,10 Age 0,00 0,01 0,01 0,14 Education Level -0,02 0,13 -0,01 -0,18 Compatibility 0,27 0,11 0,16* 2,36 Relative advantage 0,24 0,11 0,15* 2,20 Trialability 0,34 0,11 0,21** 3,04 Complexity -0,18 0,11 -0,11 -1,61 Observability 0,15 0,11 0,09 1,34 Information channel 0,25 0,21 0,08 1,19

Note. Statistical significance: *p <.05; **p <.01; ***p <.001

More research is needed to obtain more data about the different variables. Particularly the variables age, education level, complexity and observability since these did not provide significant results. It should be noted that the Cronbach’s alpha of trialability is low. The model assumed high reliability of this variable, but this is not true.

(39)

7. Discussion

In this chapter I will discuss the results from the statistical analysis and argue how the results can be related to existing theory of innovation diffusion and concepts of EEM adoption. Furthermore, I will attempt to answer the research question and demonstrate how these findings have implications in the real world, taking into consideration the limitations of this study. Additionally, future research opportunities are suggested.

7.1 Findings

I used innovation diffusion theory to explain homeowner’s intention to adopt EEM’s and showed that this theory is applicable to the context of EEM’s in the Dutch residential sector. The main findings of this study are that there is some evidence that the five perceptions of innovation characteristics are correlated to, and in part, good predictors of intention to adopt different types of EEM’s. Besides, social influencing in information channels is unlikely to have much influence over the homeowners’ perceptions of innovation characteristics and intention to adopt.

Yet, the regression models showed also that low variance was explained by the innovation characteristics. Taken together the low to moderate values in the correlation and regression analysis, it can be assumed that other factors for adoption play a more important role. The fact that this study did not correct for several barriers, thresholds and drivers that influences the homeowner’s intention to adopt, seems to limit the persuasiveness of the results. Factors like the homeowners’ financial situation (Balcombe et al., 2013) and restraining attributes of the house (Balcombe et al., 2013; Claudy et al., 2011) are factors that in some cases weigh more heavily than

(40)

an explanation of why the results of this study show low correlations in innovation characteristics. Especially the innovation’s characteristics relative advantage and compatibility were expected to give higher correlating values. Low variance explained by relative advantage and the other perceptions is uncommon. Innovation diffusion studies often report higher variances in adoption rates, in general 49 to 87 percent (Rogers, 2003, p. 206). In turn, one would expect also to see these results in adoption intention. Still, the significant results prove that the homeowners’ perceptions are to some extent predicting intention to adopt. This means that it can give guidance for future research and practitioners, since the important and less important factors for adoption intention under Dutch homeowners are unveiled. The correlation matrix and regression shows for instance that female homeowners do not perceive innovation characteristics more significantly positive or negative, but at the same time are more likely to score higher on intention to adopt. This could be the result of several underlying aspects, such as that female adopters are by nature more often associated with early adopter groups, but also, express more favour towards environmentally friendly products. This effect contrasts somewhat with findings of Diaz-Rainey & Feinberg (2015), who conclude from their own and prior research, that gender is not a reliable indicator of adoption. Regardless of its cause, practitioners in the field of energy efficiency could use this information to target their marketing efforts towards female homeowners.

For the moderating effect of the communication channel, no satisfactory results were found. The regression model that included the information channel did not provide significant results. Though, the different values of the means and standard deviation did reveal that some variables display different outcomes under the two communication channel variations. Also the intention to adopt the four selected

(41)

EEM’s was plotted and suggested that there is a slight upward and converging trend towards adoption intention when information is provided in a social manner. A rather unexpected result is that the perception of trialability correlated negatively to higher socialness of the information channel. We can also see this from the t-tests of the sample means, where the outcome of the air-sourced heat-pump was the only significant test with a ,87 lower average value. When interpreting the different results for the moderating effect, we have to be very modest and prudent. Yet, it is promising that with the relatively low sample size significant effects could be found, leaving opportunities for future research to deeper explore the effect of social influence on the diffusion process of EEM’s. Why for instance does intention to adopt seems to be somewhat positively effected by social information, while the value for trialability (a positive indicator of intention to adopt) is negatively correlated. It is highly likely, that there is some deeper explanation for this. We have already observed from the reliability test and the factor analysis that the used measure ‘perception of trialability’ is weak and too much interrelated with the other innovation characteristics. From own interpretation, the adapted trialability factor from the factor analysis in the regression analysis measured also some sort of perception of uncertainty and ignorance about the innovation’s purpose and outcome. Potentially, for the more complex and new products, such as heat-pumps, expert information can be a more effective way of stimulating the diffusion process among homeowners.

The question remains, how will these results be helpful in finding ways to stimulate EEM adoption among homeowners in the Netherlands. The goal of this study is to offer methods to effectively provide homeowners with information to close the energy

(42)

information is communicated is affecting the adoption process, but to what extent remains uncertain. Nevertheless, governments and marketers in the relevant sectors could see this as a starting point to discover opportunities that are under-examined. The results already hinted that social information provision is not necessarily improving homeowners’ perceptions of the innovation, but does positively affect the intention to adopt. This is rather counterintuitive, but could be explained by the fact that social information provision could increase EEM’s familiarity and function as a more appealing approach to inform homeowners about EEM’s. Exploring opportunities for more energy efficient housing is known as something that homeowners are not really interested in, but social information could then function as a motivating trigger to start (or further) explore relevant products. Acquiring this new knowledge about innovative products might not immediately affect perceptions in the desired way, but does trigger someone’s awareness to potentially adopt the innovation at a later stage. Perhaps, a combination of non-social information to positively affect perceptions of innovation characteristics and social information to trigger the adoption intention will successfully increase uptake. Personal stories (testimonials) of successful instalments of different EEM’s, accompanied with expert information could be shared online on social media and government websites to assist in the knowledge and persuasion stages of diffusion theory.

This study has shown several trends that are to a large extent in line with the expectations that were outlined in the hypotheses. As anticipated in H1a, the correlation matrix (table 4) shows that the four respective innovation characteristics are positively related and that for complexity (H1b) a negative correlation is present. All perceptions of innovation characteristics show a weak but significant effect. The

(43)

adapted factors from the factor analysis that were used in the regression model (table 5) show that relative advantage, compatibility and trialability significantly predict the outcome of the intention to adopt. The variables observability and complexity do not statistically predict the outcome of intention to adopt. Thus, the hypotheses are not rejected, but with only 21% of the variance predicted by the model and the variables observability and complexity being insignificant, the hypotheses H1a and H2b can also only partially be proven. The hypotheses H2a, H2b, and H2c, which predicted that the socialness of the communication channel could have a significant impact on the homeowners’ perception of innovation characteristics and intention to adopt the innovation concerned, are also not accepted. In fact, trialability showed lower mean values when exposed to information that was provided in a social manner. Therefore H2c is rejected, because an opposite effect was expected and the other characteristics did not reveal significant results. The moderating effect could not be statistically proven, which is probably due to the low sample size of this study. Though, hypotheses H2a and H2b are not rejected, since minor positive effects towards adoption could be observed and thus, the probability that social influence can be helpful in increasing adoption is still present.

7.2 Limitations

It should be noted that with the research methods that were chosen in this study, limitations occur. Surveys tend to overstate intentions (Holak, 1988; Auger & Devinney, 2007) and because of the self-report method used in this study, the possibility of the existence of the common method bias and the social desirability bias are present. Topics like sustainability that are perceived as positive in society might cause respondents to respond to questions in a socially desirable manner. However,

The effect of social influence in information channels on homeowners’ perceptions of energy efficiency measures