Scholarly article on topic 'Willingness to participate in direct load control: The role of consumer distrust'

Willingness to participate in direct load control: The role of consumer distrust Academic research paper on "Social and economic geography"

CC BY-NC-ND
0
0
Share paper
Academic journal
Applied Energy
OECD Field of science
Keywords
{"Demand management" / "Load control" / "Peak demand" / Distrust / Trust}

Abstract of research paper on Social and economic geography, author of scientific article — Karen Stenner, Elisha R. Frederiks, Elizabeth V. Hobman, Stephanie Cook

Abstract Addressing the challenge of peak demand is a major priority for energy utilities, regulators and policymakers worldwide. Against this backdrop, residential demand management solutions – including direct load control technology that allows utilities to turn specific household appliances on and off during peak periods – are becoming increasingly important. While such technology has been available for decades, acceptance and adoption among residential consumers has not always kept pace. Why is this so? Drawing on key principles from psychology and behavioural economics, we propose that consumer distrust can play a significant role in the uptake of demand management solutions. As part of a large field study, a survey-experiment was conducted to investigate householders’ willingness to participate in a direct load control program offered by an Australian energy company. To specifically examine the relationship between self-reported distrust and willingness to participate, and how this relationship might be influenced, the survey included an unobtrusive experimental manipulation: a simple two-sentence message designed to rebuild consumer trust and confidence in the utility was conveyed to a randomly-selected subsample of participants. Results suggested that participants’ self-professed distrust in the utility was associated with significantly lower willingness to register for the DLC program. This unwillingness was modestly reduced for those participants who received the trust-restoring message upfront. Together, these results suggest that distrust may serve as an important decision-making heuristic used by consumers when choosing whether to accept new demand management technology and services. Implications for future research and practice are discussed.

Academic research paper on topic "Willingness to participate in direct load control: The role of consumer distrust"

Contents lists available at ScienceDirect

Applied Energy

journal homepage: www.elsevier.com/locate/apenergy

Willingness to participate in direct load control: The role of consumer distrust

Karen Stennera'*, Elisha R. Frederiks a, Elizabeth V. Hobmana, Stephanie Cook1

a CSIRO Adaptive Urban & Social Systems Program, Ecosciences Precinct, 41 Boggo Road, Dutton Park, Qld 4102, Australia2

HIGHLIGHTS

• Consumers use trust and distrust as decision heuristics to help guide behaviour.

• Distrust may reduce willingness to participate in direct load control programs.

• Efforts by a utility to regain customer trust may increase willingness to participate.

• Using mixed methods, we conduct a survey-experiment to examine these issues.

ARTICLE INFO ABSTRACT

Addressing the challenge of peak demand is a major priority for energy utilities, regulators and policymakers worldwide. Against this backdrop, residential demand management solutions - including direct load control technology that allows utilities to turn specific household appliances on and off during peak periods - are becoming increasingly important. While such technology has been available for decades, acceptance and adoption among residential consumers has not always kept pace. Why is this so? Drawing on key principles from psychology and behavioural economics, we propose that consumer distrust can play a significant role in the uptake of demand management solutions. As part of a large field study, a survey-experiment was conducted to investigate householders' willingness to participate in a direct load control program offered by an Australian energy company. To specifically examine the relationship between self-reported distrust and willingness to participate, and how this relationship might be influenced, the survey included an unobtrusive experimental manipulation: a simple two-sentence message designed to rebuild consumer trust and confidence in the utility was conveyed to a randomly-selected subsample of participants. Results suggested that participants' self-professed distrust in the utility was associated with significantly lower willingness to register for the DLC program. This unwillingness was modestly reduced for those participants who received the trust-restoring message upfront. Together, these results suggest that distrust may serve as an important decision-making heuristic used by consumers when choosing whether to accept new demand management technology and services. Implications for future research and practice are discussed.

© 2016 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4XI/).

CrossMark

Article history:

Received 2 March 2016

Received in revised form 24 October 2016

Accepted 25 October 2016

Keywords:

Demand management

Load control

Peak demand

Distrust

1. Introduction

Peak demand - the daily and/or seasonal spikes in consumer demand for electricity - plays an important role in the costs of

* Corresponding author.

E-mail addresses: karen.stenner@csiroalumni.org.au (K. Stenner), elisha.freder-iks@csiro.au (E.R Frederiks), elizabeth.v.hobman@csiro.au (E.V. Hobman), Stepha-nie.Cook@uqconnect.edu.au (S. Cook).

1 Ms Cook was a Research Officer at the CSIRO at the time the research study was conducted.

2 Note: CSIRO refers to the Commonwealth Scientific and Industrial Research Organisation.

electricity generation and supply. Demand-side management (DSM) solutions to curb peak demand have therefore gained significant attention among industry stakeholders worldwide, as they offer an effective means of reducing future investment in costly network infrastructure that is specifically built to meet maximum demand levels. DSM is an overarching term that describes an increasingly diversified range of activities, but can be broadly defined as ''a utility action that reduces or curtails end-use equipment or processes [and] is often used in order to reduce consumer load during peak demand and/or in times of supply constraint" [1]. It includes everything that targets the demand side of an energy system [2], ranging from brief curtailment of energy usage via load

http://dx.doi.org/10.1016/j.apenergy.2016.10.099 0306-2619/® 2016 The Author(s). Published by Elsevier Ltd.

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

management systems through to more ongoing or permanent improvements in efficiency via new energy-saving technology.3 DSM activities can be classified into several categories [for a conceptual review and taxonomy, see 2], but are broadly understood to include both energy efficiency (e.g., permanently reducing demand via more efficient appliances) and demand response (e.g., reactive or preventative measures to reduce, stabilise or shift demand, including incentive-based programs like load control and price-based measures like dynamic tariffs) [5-7].4

1.1. Value and untapped potential of residential DSM

Although the total level of load reduction potential is highest for industrial and commercial consumers [8,9], the potential benefits that can be achieved by optimising DSM participation among residential consumers cannot be understated.5 DSM can be particularly impactful for households because contrary to industrial and commercial loads that are often application-specific, residential demand is primarily shaped by a small number of energy-intensive domestic appliances that have widespread market penetration and regular use [14]. The number of individual end-users also tends to be higher in the residential sector, so there is sizable scope for improving DSM participation rates en masse. It is perhaps unsurprising, therefore, that utilities are increasingly promoting consumer uptake of automated demand management solutions that directly target the discretionary load of domestic electrical appliances in residential dwellings, with the aim of reducing usage during peak periods and/or shifting usage to off-peak times. Such solutions are seen to offer a highly replicable means of controlling peak demand across the residential sector, with significant potential for widespread uptake and usage.

Direct load control (DLC) devices are one such solution. Broadly speaking, DLC technology allows utilities to remotely manage demand for electricity by directly modifying the operation of end-use devices - typically air conditioners, pool pumps and electric hot water systems. Ordinarily, DLC programs involve a utility or system operator installing equipment (e.g., radio-controlled device, known as a 'remote appliance controller') that allows them to switch specific appliances on and off for a short time during peak periods and critical events [1,15]. In return for participating, consumers are usually rewarded by way of a financial incentive such as a one-off signup payment, recurring annual payment, ongoing electricity bill discounts, or free hardware installation. DLC programs have been around since the early 1970s [16] and are arguably the most common type of demand response program [7]. This form of DSM is highly attractive for networks and system operators by assisting with more accurate planning of future investments in capacity [17-19]. From a network's perspective, the ability to control load during critical peak periods - particularly for appliances that make the largest contribution to residential peak demand - allows for more reliable demand forecasts at the localised level, by providing greater certainty over the amount, timing and location of potential energy savings. For these solutions

3 For detailed reviews of recent demand-side developments, and the benefits and challenges of demand response, see [3,4].

4 More specifically, EE measures are designed to encourage consumers to reduce overall energy usage (i.e., 'load reduction') via efficiency actions (e.g., one-off behaviours such as purchasing new energy-efficient appliances, installing insulation/retrofits, etc.) and/or curtailment actions (e.g., everyday behaviours to conserve energy, such as switching off appliances when not in use, adjusting thermostat levels, etc.). DR measures are designed to transfer consumer load from peak to off-peak periods (i.e., 'load shifting'), usually by rewarding consumers for reducing electricity demand during certain times.

5 DSM in the industrial and commercial sectors is not the focus of this paper.

However, there is a growing body of literature available for readers who are interested in this area (e.g. [10-13]).

to yield optimal long-term benefits, however, it is important for consumers to respond positively - that is, they must accept and adopt the new DSM technology, and willingly participate in load control programs that are offered to them. Clearly then, consumer decision-making and behaviour (the 'human' aspects) - in conjunction with various contextual factors and the social structure in which people operate - play an important role in determining the effectiveness of new technology and policy initiatives designed to change household energy usage [20-22].

Despite offering a range of potential benefits for consumers, utilities and networks alike, and despite ongoing technological advances with the variety and capability of DLC devices, widespread consumer uptake and usage remains surprisingly low. In the United States, for instance, a 2012 survey by the Federal Energy Regulatory Commission reported that customer enrolments in DLC programs ranged from a mere 0.11% in the Texas Reliability Entity region, up to just 14.54% in the Florida Reliability Coordinating Council region [23].6 In terms of residential customers more specifically, although the largest DLC programs have enrolled hundreds of thousands of participants, this equates to a very small percentage of the overall population. From a consumer psychology perspective, householders may be unwilling to participate for myriad reasons such as a perceived lack of control, concerns over disruptions to one's lifestyle or comfort, limited knowledge/awareness, and - as we argue herein - a sense of distrust and scepticism. In order for industry stakeholders to improve current rates of uptake and usage - and ultimately maximise market penetration of DSM solutions across the entire residential sector - it is therefore critically important to gain greater evidence-based insights into the specific factors that underpin consumer decision-making and behaviour around DSM. That is, we need to identify and better understand the most powerful and pervasive motives that lead people toward accepting (vs. rejecting) DSM solutions and programs - something that has received surprisingly little attention to date.

1.2. Prior research on behavioural drivers/barriers to participation

In the academic literature, evidence-based insights on the psychological drivers and barriers to consumer participation in residential DLC programs are surprisingly sparse. Although a number of DLC field trials and pilot programs have been undertaken across the globe, the results of such studies are often limited to business, industry and government reports and conference papers rather than peer-reviewed journals [e.g., 24,25-32]. Many studies also tend to focus more heavily on the technical, economic and market-based drivers and barriers for growth in the residential demand response market [7], with comparatively less focus on examining the psychological factors (i.e., cognitions, emotions, behaviours) that shape consumer responses. Furthermore, while recent years have seen greater recognition of the 'human' aspects, in such cases this is often more exploratory research that is not designed in a way to allow causal conclusions to be drawn. For example, very few scientifically rigorous studies have been conducted to identify the causal factors (predictors) and contingencies (moderators) that explain why people respond to DSM solutions like DLC in a certain way. As such, some questions remain over exactly what motivates consumers to participate (or not) in DLC programs, as well as how, when, where, why and for whom such motivations apply.

To confidently answer such questions about causality, a robust experimental design is required [for further discussion on the value of experimentation, and the criticality of randomised

6 Note that these figures pertain to all retail customers, i.e., any purchaser of energy that consumes electricity for residential, commercial or industrial use, or a variety of other end-uses [23].

controlled trials for evaluating the 'impact' of an intervention on behaviour, see 33,34-39]. Scholars are increasingly recognising that ''customer behaviour is unstable, changeable and unpredictable" [40, p. 25], and recent modelling and simulation studies have made significant progress by considering the complexities of human behaviour [for examples, see 40,41]. But robust scientific experiments actually testing these insights in the field with real-world consumers have been slow to emerge.7 Similarly, while a growing number of conceptual reviews, qualitative studies, self-report surveys, and even small-scale pilots have been conducted to explore consumer preferences, attitudes and experiences relevant to DSM [for examples, see 31,43-46], surprisingly few experiments - specifically in the form of randomised controlled trials - appear in the literature.8 This gap highlights the critical need for more empirical work to specifically test the proposed causal factors that explain consumer responses to DSM technologies such as DLC.

Despite the lack of extensive experimental research, more general insights from the social sciences - particularly the fields of psychology and behavioural economics - may help elucidate some of the key factors that underpin consumer decision-making around household energy consumption, including specific DSM actions and technology [48-53]. In particular, and as explained further in the next section, theory and research on human decision-making suggests that one important factor shaping consumer uptake and usage of a new product or service is trust and/or distrust - not only in the product or service itself (e.g., DSM technology and programs), but also in the entity (e.g., utility) offering it. In this paper, we examine this hypothesis by presenting some relevant results from a larger multi-experiment survey that investigated consumers' willingness to participate in an automatic DSM initiative being offered by an Australian utility. This was a DLC program for residential electric hot water storage systems, which would allow the utility to remotely turn off a customer's hot water system for a short time during peak periods as a means of managing peak demand. Using a mixed methods approach, we systematically examine the association between consumer distrust and self-reported willingness to participate in the program. We test the impact of a simple trust-restoring message designed to acknowledge while also allaying any existing feelings of distrust in the utility offering the program.

1.3. Influence of trust and distrust on consumer decision-making

In daily life, people face a seemingly endless array of choices and decisions. Yet human cognition is limited in its capacity to process the vast amount and complexity of stimuli that characterise everyday situations. In most cases, it is impractical or impossible for people to engage in highly systematic, rational information-processing. Instead, people tend to rely on cognitive simplifying devices known as 'heuristics' - that is, mental 'shortcuts' and 'rules-of-thumb' that help reduce cognitive overload and facilitate more rapid information-processing [54-58]. Relying on simple heuristics is especially pronounced when people face high levels of complexity, choice or uncertainty - all of which may place heavy

7 For a recent review of the methodological problems and design weaknesses that are common in electricity usage pilot studies, including studies of automated technologies that control energy use, see [42].

8 One exception is a recent study by Fell et al. [47], which involved a between-subjects survey experiment with a representative sample of British bill-payers. Results revealed that a tariff permitting DLC of home heating (in return for a reduced flat rate for electricity) was deemed more acceptable than any other static or dynamic time of use tariff presented (and regardless of automation). The DLC tariff was rated more favourably in terms of providing a general sense of control (regarding comfort, timing and autonomy), as well as perceived usefulness and ease of use. According to Fell et al., this suggests that when operated within tight bounds and with override ability, DLC is acceptable 'in principle' to many people.

demands on human cognition. In such situations, using heuristics allows us to bypass more intensive information-processing and conscious deliberation, such that decision-making becomes simpler, easier and faster. In the residential energy domain, for example, a range of cognitive biases, heuristics and psychological factors may influence the decision making and behaviour of consumers and households (for reviews, see [48,59]).

While a plethora of well-known heuristics are cited in the literature [54-58], of particular importance to understanding how consumers will respond to DSM is the use of trust and distrust as decision-making tools - that is, one's perception of trust and credibility may serve as a mental 'shortcut' that eases the cognitive load of making a decision, particularly in the face of complexity, risk and uncertainty [60-63]. Research suggests that trust and credibility may be important determinants of residential energy-related choices and behaviour [21,64-66], as well as acceptance of energy innovation and technology [67,68]. Accordingly, we contend that the extent to which consumers trust or distrust a new DLC technology or program and (importantly) the utility offering it may influence their acceptance and adoption of the DLC initiative.

Despite some inconsistency in the literature over the conceptualisation of trust and distrust, most theorists agree that they are separate constructs with different characteristics and determinants [see 69-71], rather than just opposites on the same continuum. From a multidisciplinary perspective, trust has been defined as ''a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behaviour of another" [72, p.395]. It reflects favourable expectations regarding another's conduct, and is associated with positive emotional reactions such as hope, confidence and assurance [71]. Conversely, distrust reflects unfavourable expectations and involves negative reactions such as suspicion, wariness and fear [73]. Distrust has been defined as ''a lack of confidence in the other, a concern that he or she may act so as to harm us, that he or she does not care about our welfare, intends to act harmfully, will not abide by basic moral norms, or is hostile towards us" [73, p. 34]. In this regard, trust is conceptually distinct from (albeit correlated with) distrust. Yet there are still some underlying similarities between the concepts. As Lewicki, McAllister and Bies [71] note, ''both trust and distrust involve movements toward certainty: trust concerning expectations of things hoped for and distrust concerning expectations of things feared" (p. 439). An inherent part of this notion of movement towards certainty is risk

- the potential of loss (vs. gain) in situations of uncertainty. Risk is considered a necessary but insufficient condition for building trust

- without risk neither trust or distrust would arise, but it is not the only prerequisite [74]. When an outcome is entirely certain, for instance, trust is unnecessary because people can make decisions and act with absolute certainty of the consequences, i.e., one has full confidence in achieving the outcome that is expected.

Extensive research suggests that trust is used as a mechanism for reducing complexity and intensive cognitive processing [61,71,75-77], particularly in uncertain or 'risky' situations where people must make cost-benefit appraisals and evaluate the likelihood of loss (vs. gains). In such situations, perceptions of credibility and trust can serve as a barometer of 'risk' when choosing a particular course of action. This use of trust as a simple heuristic extends across a range of contexts, including the domain of consumer choices and behaviour. When deciding whether to purchase a new product or request a new service, for example, residential energy consumers may think about the potential losses, costs and risks of shifting from the status quo. This might involve consideration not only of the time, effort and money involved, but also the potential impacts on personal wellbeing (health, safety, happiness), lifestyle (comfort, habits, routine), and the environment

(carbon emissions, ecological sustainability), among other factors (for a review, see [48]). Perceptions of credibility and trust might strongly influence how people evaluate such risks (vs. benefits) and in turn, what choices and actions are taken.9 A new offering is unlikely to be accepted if people view the product or service itself, or the entity or information promoting it, as untrustworthy, dishonest, unfair or lacking credibility. Conversely, if these are perceived to be credible, plausible, genuine and trustworthy, then consumer acceptance is likely to be enhanced.

In support of this notion, many studies have found that trust can influence the risk-benefit appraisals and perceived acceptability of new technology [61,78]. This includes, but is not limited to, studies on carbon dioxide capture and storage [79-81], new hydrogen systems [82], gene technology [83,84], nanotechnology foods [85], pesticides, nuclear power and artificial sweetener [62] - as well as various other hazards [75,86]. For example, Siegrist [83] found that trust in institutions responsible for using gene technology had a positive impact on perceived benefit (and a negative impact on perceived risk) of this technology, with these benefit/ risk perceptions subsequently affecting one's acceptance of biotechnology. Another study by Siegrist and Cvetkovich [75] found that personal knowledge of a hazardous activity or technology also plays an important role. When self-assessed knowledge about an activity or technology was low, social trust in the authority regulating the activity/technology was strongly related to judgements of its risks and benefits. In contrast, this relationship did not hold for activities and technologies about which people were already knowledgeable. A large body of risk research also shows that peoples' trust in organisations that manage hazardous activities and complex technologies partly depends on the extent to which these organisations are seen as objective and unbiased; open, honest and fair; consistent and predictable; faithful and committed; concerned with the public interest; and competent with relevant technical expertise [60,63,87,88]. Recently, Terwell et al. [79,81,89,90] proposed a causal chain of trust to explain how public acceptance of new technology is shaped by perceptions of two broad types of trust: competence-based trust (perceived organisational expertise and experience) and integrity-based trust (perceived organisational honesty, openness, concern for public interests).

In the energy domain, the perceived credibility of information (e.g., energy-saving messages and consumer-focused communication) has even been found to impact householders' requests for energy-saving information and the actual consumption of electricity. For example, a field experiment by Craig and McCann [65] found that the source of consumer communication had a noticeable influence on energy conservation behaviour. Specifically, messages originating from a high credibility source (e.g., public service commission) generated significantly more requests for energy conservation information, and greater conservation of electricity, than the same message originating from a less credible source. The authors conclude that using a more credible source can enhance the effectiveness of communications advocating energy conservation. This notion is consistent with Costanzo et al.'s [91] social-psychological model of energy use behaviour, which proposes that the efficacy of informational appeals to conserve energy, as well as the adoption of a conservatory attitude, are partly influenced by

9 As noted by Huijts et al. [68], much research suggests that trust can shape the acceptability of (or intention to accept) technology indirectly, via perceived costs and risks. Greater trust tends to enhance the perception of benefits while reducing perceived costs/risks, thus boosting acceptance. Some studies model trust as a direct predictor of intention to accept. Others suggest that trust may exert its influence more indirectly, by first influencing intermediate affective states. The feelings thus generated might then go on to influence perceived risks and benefits. That is, trust may elicit more positive emotional reactions to a technology, which then lead to more favourable evaluations of its risks, costs and benefits.

the credibility of the message source.

Together, these insights have important implications for understanding consumer decision-making in response to DSM solutions - and this naturally includes DLC technology and programs, which form the focus of our paper. Based on prior research, we argue that receiving information about DLC initiatives from a more credible, trustworthy source will be associated with a more positive consumer response compared to receiving the same information from less credible, distrusted sources. In particular, we predict that consumer willingness to accept DLC will be negatively impacted by self-reported distrust of the entity (e.g., energy utility) offering the technology/program. We also argue that efforts to establish or strengthen trust - for instance, consumer-focused messages or communication to build consumer confidence - may lead to a more positive response.

1.4. The current study

To investigate our hypotheses, we conducted a randomised survey-experiment with a large sample of residential energy consumers (i.e., householders) from one state in Australia. As one part of a broader study, we aimed to examine the role of distrust in self-reported willingness to participate in a specific type of automated DSM: a direct load control program for electric hot water storage systems being offered by an energy company. The survey was informed by an earlier qualitative study (in the form of householder focus groups, and targeted domain and stakeholder analysis) that revealed some rather intense distrust of the utility among our target population of householders. For example, across eight focus groups, we found that over 8% of all comments offered by participants directly reflected self-professed personal distrust and/or observations of general public distrust of the utility. Such remarks often included explicit predictions that this distrust would undermine community interest in participating in the utility's DLC program.

Accordingly, our first aim was to investigate whether self-professed distrust in the energy company - as reflected in participants' qualitative responses to our open-ended questions - was associated with less willingness to subscribe to the program. We hypothesised that those participants who mentioned distrust of the utility would be much less likely to subscribe than those who did not express any such wariness or suspicion. Second, we aimed to test whether providing randomly selected participants with a trust-restoring message upfront -an explicit acknowledgement of community disaffection with the utility, and reassurance of the utility's commitment to rebuilding consumer support/trust -would influence their responses. We hypothesised that receiving such a message would have a significant positive impact by increasing willingness to participate in the DLC program. Finally, we aimed to compare the specific impact of the trust-restoring message for participants who did and did not report a sense of distrust. We hypothesised that the impact of this message would be greater among those who specifically mentioned distrust in the utility as a reason for not participating. Our three main hypotheses were therefore as follows:

H1. Self-professed distrust in the utility will be associated with significantly lower willingness to participate in the DLC program.

H2. Providing a trust-restoring message that explicitly acknowledges consumer distrust in the utility and provides reassurance of efforts to rebuild trust will significantly increase willingness to participate.

H3. Conveying a trust-restoring message will boost willingness to participate to a greater degree among those who single out distrust in the utility as a reason for not participating.

2. Method

2.1. Experimental design

The research questions, hypotheses and results presented in this paper are derived from a larger multi-experiment field study conducted with householders in one state of Australia. The study involved several distinct experiments (each conducted independently, with different goals in mind) embedded into a survey, examining the impact of different randomly-assigned messages and incentives on participants' self-reported willingness to subscribe to the DLC program. Each experiment incorporated its own control group and one or more alternative experimental groups ('treatments'), all randomly assigned. The focus of this paper is squarely on the first experiment, which investigated the impact of a simple two-sentence message on participants' willingness to subscribe to the DLC program. We do not provide further details on the other experiments embedded within the survey, as we focus here solely on the research questions that motivated and underpinned the design and conduct of this first experiment.10

In regard to the latter, we unobtrusively embedded an experiment within the survey in which participants were randomly assigned to either receive (treatment group) or not receive (control group) the following 'trust-restoring message' upfront, in the very first section of the survey:

'Now, you may be aware that some people out there in the community don't have a great deal of trust in [the energy company]. [The company] is trying hard to rebuild community support and is committed to providing customers with high quality service at the lowest possible cost'.

This message incorporated a public acknowledgement of existing community disaffection coupled with a reassurance from the utility aimed at restoring consumer trust. The critical behavioural outcome of interest, which was expected to vary across the treatment (message) and control (no message) groups, would be subsequent self-reported willingness to sign up for the DLC program. Participants11 who continued on with the survey were provided with a brief description of the utility's proposed DLC program. Here they were advised that the company would soon be offering customers a completely free and voluntary program. The only requirement would be that they allow the company to install a new meter in their homes that would ''automatically switch off your hot water system for a short time during peak demand periods, when everyone's trying to use electricity at the same time". Participants were advised that ''it would all be pre-scheduled, so you'd never have to lift a finger, or give it any thought". They were also reassured that their hot water would remain hot, despite their system being switched off for a short period.12

10 A comprehensive analysis of the survey's qualitative data (i.e., participants' open-ended comments) is presented in [92] and available upon request.

11 Other than those who completed only the two-minute version of the survey

12 As noted earlier, several other experiments were embedded within the same survey to test the impact of message framing and incentives related to the DLC

program. Thus, randomly selected participants were also advised about monetary rewards or other material benefits, given various reassurances (apart from this one designed to restore trust), and/or presented with different explanations of peak demand. But all of these distinct experiments were independently randomly assigned and thus their effects are not confounded.

To gauge our key behavioural outcome of interest, we wanted a measure of 'willingness to participate' that would reflect as closely as possible participants' likelihood of subscribing to the program, without actually asking them to officially register, as the latter was prohibited both by the research ethics committee and the survey provider.13 Accordingly, participants were first presented with the question: ''Would you register for the [DLC program] once it begins?" They were then asked to think about a probability scale ranging from 0% (meaning ''There's no chance at all I'd sign up") through 50% (''There's a 50/50 chance I'd sign up") to 100% (''I'd definitely sign up"). Participants were asked to indicate on this scale ''.. .how likely is it that you would actually sign up for a program like the [name of DLC program]?" The scale increased in 5% increments from 0% to 100%, so participants' responses were ultimately dispersed in a finely graded manner across 21 points reflecting varying levels of interest in the program. This constitutes our dependent (outcome) variable for the analysis.

After indicating their willingness to register for the program, participants were immediately asked to describe the factors influencing their choice by answering the open-ended question: ''In deciding whether you might want to participate in this program, what influenced your decision? What would be your main reasons for participating or not participating?" Through this open-ended question we sought to uncover some of the reasons that (at least according to participants' own reports) might serve to boost or hinder participation in the DLC program. The distrust in the energy company expressed by some participants in response to this question constitutes (alongside the experimental manipulation of reassurance that we described earlier) one of the two explanatory variables in our analysis.

Finally, the survey went on to ask (of those willing to complete the long form of the questionnaire) a range of other open-ended and forced-choice questions covering consumers' current energy-related practices and attitudes (see Appendix). These variables are not deployed in the current analysis.

2.2. Participant recruitment

A total of 1499 householders from one state in Australia participated in the study. To be eligible, potential participants had to be residents 18 years or over, and were constrained overall to be representative of the state's population on age and gender. Participant recruitment was managed by a third-party research agency (survey provider) using three different channels, as outlined below.

2.2.1. Face-to-Face recruitment

Field interviewers conducted face-to-face surveys at three busy shopping districts within the capital city of the state in question. Where the quotas for gender and age were still open and there was sufficient pedestrian traffic, every sixth person who passed by was approached and invited to participate. Otherwise, participants were selected based on the 'next available' person meeting the prevailing quotas. A total of 553 participants agreed to complete the survey via this face-to-face method, with the interviewer reading out questions to the participant and then inputting their answers into an online survey via an iPad.

2.2.2. Telephone recruitment

Random digit dialling was used to select participants to complete the survey via telephone. Of the 5336 randomly generated telephone numbers that were called, 2561 individuals were spoken to. Of these, 439 were ineligible for the study (due to age and gen-

13 This measure is similar to what Aronson and colleagues [93] have referred to as a 'behavioroid' measure, which is designed to reflect ''subjects' commitment to perform a behavior, without actually making them carry it out" [93; p. 271].

der quotas), 1654 refused to participate, and 468 agreed to complete the survey - representing a response rate of around 22%.

2.2.3. Online recruitment

Two panels of householders (who had given the survey provider their prior consent to participate in research studies) were used in parallel to recruit participants to the online survey. Standard procedures to eliminate duplicate responses were applied across the panels, based on email and IP addresses. Invitations were emailed to a total of 3130 people, of whom 658 launched the survey and 478 completed it. This represent a response rate of just over 15%.

Overall, of the 1499 householders who completed the survey across one of the three modes (face-to-face, telephone or online), 9% (n = 134) were identified as ineligible to participate in the DLC program due to having gas or solar hot water systems, being entirely ''off the grid", and/or being in the process of relocating homes and unsure of their new arrangements. Unless otherwise stated, the results presented here pertain only to eligible participants (n = 1365), and in the case of analyses utilising the qualitative data generated by our open-ended items, only those who answered those particular questions (n = 1168).

Based on socio-demographic data collected from those completing the 20-minute survey (n = 838), the final sample seemed reasonably representative of the state's population, as gauged against a range of individual, family and household variables.14 While cautioning that comprehensive socio-demographic data were not collected for those who completed shorter versions of the survey, to the extent that we were able to judge,15 it appears there were no marked differences between those completing the longer and shorter versions.

2.3. Survey administration

There were three different modes of survey delivery (face-to-face, telephone and online), and three different lengths of survey (2 minute, 10 minute and 20 minute) requiring widely varying commitment from participants.16 Our goal in designing the survey in this fashion - allowing respondents to select for themselves the time they were willing to invest, and administering the survey via three different channels - was to attract as broad and representative a slice of the customer base as possible, including capturing those

14 We say ''reasonably" representative mostly because there were certainly more females than males (58% vs. 42%), and the sample did over-represent middle-aged people aged 30 to 64 years (80% of the sample). Participants ranged from 18 to 85 years, with an average age of 47 years. High school graduates were over-represented (53% in sample vs. 36% in population), while university qualifications (Bachelor degree or higher) were slightly under-represented (23% in sample vs. 28% in population). Single-person households were rarer than the norm (15% vs. 28%) while those with three or more residents were more common than expected (46% vs. 36%). Close to the population norm (86%), the vast majority of our participants (83%) lived in detached houses. Importantly, there was also a very close match between sample and population in home ownership (just over 70% for both) and also household income where (in each case) about 15% of people received less than $400 per week, about 35% received $400 to $999, 32% received $1000 to $1999, and 13% $2000 or more per week (~5% missing).

15 By attributing to each individual the socio-demographic attributes associated with their postcode, and then looking for any notable socio-demographic variation between those undertaking different survey lengths.

16 The 2 minute version of the survey was designed (with its minimal commitment) to attract reluctant participants that might then be enticed (by additional incentives offered once in progress) to complete the longer forms of the survey. This brief version covered just a handful of simple items such as the participant's postcode, an estimate of their last electricity bill amount, and the types of energy used in the household. The 10 minute survey went on to describe the DLC program and collect the key variables (willingness to participate, and purported reasons for this choice). The 20 minute survey went even further by asking about lifestyles, energy practices and attitudes, and basic socio-demographic attributes. Although our key dependent variable (willingness to subscribe) was only presented in the 10 and 20 minute surveys, the vast majority of our sample (87%; n=1,310) opted to go to these lengths.

energy consumers who might be intimidated by and under-represented in (especially longer and more complex) surveys. This would maximize external validity and thus our ability to generalise from the survey findings to the relevant population. Time commitment was voluntarily selected by the participant at the very outset and could be re-considered as they progressed through the survey items, with additional incentives offered at different stages for continuing on. In return for taking part, participants earned one or more entries into a random cash prize draw, with the number of entries contingent on the length of survey they agreed to complete: one entry for the 2 minute survey, five entries for the 10 minute survey, or 10 entries for the 20 minute survey.

In terms of delivery modes, about 37% did the survey face-to-face, while around 32% completed it online and 31% via telephone. More than half the sample (56%) finished the (longest) 20 minute version of the survey, and just under a third (31%) completed the 10 minute version, while only about one in eight participants (13%) opted for the (shortest) 2 minute survey. A two-way tabulation revealed significant differences in survey length across the different survey modes (v2(4) = 460.83, p< 0.001). The 20 minute survey was very heavily favoured by online participants (91%), while in the telephone mode, the 10 minute (49%) and 20 minute versions (46%) proved nearly equally popular. The face-to-face delivery mode had the most diverse distribution, with participants somewhat more evenly spread across the 2 minute (26%), 10 minute (40%), and 20 minute (34%) versions.

Surveys were completed for all three delivery modes (online, telephone and face-to-face) over a four week period in late 2012. Data were initially recorded in three separate files, according to survey mode. Each of the three data files was cleaned (removing test cases and duplicates) and then merged to form a single data file for statistical analysis.

3. Results

Preliminary analyses revealed that the average likelihood of subscribing to the DLC program (as indicated on the 0-100% rating scale, described earlier) was around 63% among eligible participants. Note, however, we do not expect that this figure (63%) represents the likely uptake rate among the broader target population, given that it is based only on an (inevitably, somewhat selective) sample of consumers who were willing to take part in our survey. Bear in mind our primary focus here is on some potential determinants of uptake, rather than the likely level of uptake per se. If we wanted to arrive at some realistic estimate of the broader consumer uptake of such a scheme upon roll-out, we would need to take into account not just willingness to subscribe among those to whom we are able to pitch the DLC scheme (here, 63%), but also what percentage of the original 'targets' even allow us to make that pitch in the first place (roughly indexed in the present case by our survey response rate, which hovers around 20% of the starting sample).

Just on this survey evidence, then, we might estimate a likely uptake rate of around 13% (that is, 0.2 x 63). But we imagine that in any real world roll-out, the percentage of customers who would 'allow' the DLC sales pitch to be made (e.g., by considering some billing insert accompanying their power bill) might well be higher than the (only roughly analogous) 20% response rate we achieved in our survey. After all, that would be a real and immediately available DLC offer, mailed out to actual customers of the utility along with their regular billing (or perhaps intensively marketed in their neighborhoods door-to-door). This is bound to receive greater attention and consideration from the 'targets' than some future offering in an unsolicited survey addressed only to a broad slice

of the adult population, who are not all customers of the utility, or even bill-paying heads of households.

In any case, we reiterate that pinpointing the exact level of uptake of such a scheme is only of passing interest in the current investigation. Here our focus is squarely on some likely determinants of customer uptake, and specifically, the potential role played by consumer distrust. Our survey-experiment is purposely designed, and well-equipped to answer that question. The aforementioned 'likelihood of subscribing' scale constitutes our dependent (outcome) variable for the analyses to follow. The key question is: to what extent did trust/distrust of the utility - both self-reported by the respondents, and experimentally manipulated by our randomly assigned trust-restoring message - influence that professed likelihood of subscribing to the DLC program?

3.1. Association between distrust and unwillingness to participate

To test our hypotheses, we first systematically reviewed and coded participants' responses to our open-ended questions, specifically looking for (self-reported) ''main reasons for participating or not participating" in the DLC program, i.e., for the factors that respondents themselves (at least) believed ''influenced [their] decision" about ''whether [they] might want to participate in this program". We followed a two-step process to transform all the qualitative data generated by our open-ended questions into dichotomous categories capable of being incorporated as explanatory variables in our quantitative analysis of willingness to subscribe. First, after reviewing the commentary, a comprehensive coding scheme was developed covering the range of reasons offered by respondents for their decision on whether to participate in the DLC program. Second, two of the authors used this coding scheme to code (independently) all open-ended responses. To achieve parsimony and improve interpretability, similar codes were combined, and only codes that were mentioned by at least 2% of the original sample were included in the analysis.17 The codes were transformed into a series of dummy variables (0 = did not mention this factor, 1 = did mention this factor) and entered as predictors (of willingness to subscribe to the program) in ordinary least squares (OLS) regression analysis.

Participants cited a range of reasons for not subscribing to the DLC program.18 The relative importance of these apparent barriers can be determined by investigating the associated reduction in the probability of participating, i.e., the decrease evident in their (self-reported) likelihood of subscribing if that particular barrier was mentioned by the respondent (vs. not mentioned). We designate factors 'important' if they were associated with a statistically significant decrease in the likelihood of participating in the DLC program. While a range of rationales and explanations for not participating were reported, distrust of the utility did seem to play a particularly pow-

17 This 'rule of thumb' (for the cut-off) involving 2% of the sample (here, 30 cases) was developed from experience with multiple regression analyses involving dummy variables formed from multi-theme commentary generating widely dispersed categorical data. In essence, any category capturing less than 2% (or alternatively, more than 98%) of the cases effectively has insufficient variance to permit any sensible analysis. That is to say, the dummy variable will simply have too many cases (too great a proportion) scoring '0', where almost no-one mentioned that (potentially) explanatory factor (or alternatively, too many scoring '1', where nearly everyone mentioned the factor) to allow any meaningful analysis of how variation ('0' or '1') in that factor (explanatory variable) might explain variation in the outcome (dependent

18 Full details of other reasons associated with significant reduction in likelihood of

subscribing can be found in [92]. These included general disinterest; insufficient material rewards; already limiting energy use/ performing manual DSM; concern about disruption to lifestyle/routine; inconvenience/intrusion of program/installation; fear of negative impacts on system/energy consumption; being a renter; wanting more consumer testimonials/evidence of program success; concerns about financial costs; and wanting more time/information to make a decision.

erful role.19 Our analyses indicate that self-professed distrust of the utility was associated with a significantly reduced willingness to participate. As reported in Table 1, and depicted in Fig. 1, those indicating they would not participate due to distrust of the utility, suspicion around the program's integrity and/or prior negative experiences with the utility (2.35% of participants mentioned these factors) were significantly less interested (p < 0.001) in the DLC program, being about 45 percentage points less likely to subscribe (likelihood of participating = 19%) than those who did not specify any such mistrust or suspicion (likelihood of participating = 64%). Thus, in support of our first hypothesis, our results suggest that self-professed distrust of the utility was significantly associated with lesser willingness to subscribe to the DLC program.

Some examples of comments that reflect the apparent influence of distrust include the following:

• The past reputation of the company has influenced my decision.

• It's just a scheme to make more money.

• The idea is a scam. It does nothing to really save money.

• I see it as another way for [the utility] to make more money.

• [The company] will rip us off... [they] don't care about the common person, just the almighty dollar.

• I have no faith in [the company].

Interestingly, respondents' qualitative comments - unconstrained, voluntary responses to our open-ended questions about the reasons for their decision - were far more powerful in predicting unwillingness to subscribe to the DLC program than the quantitative responses from two closed-ended questions we had also inserted into the survey. At the end of the survey respondents had been asked to indicate the extent to which they agreed or disagreed with each of two statements: ''I don't really trust [the company]" and ''I feel [the company] is generally trustworthy". Participants' distrust as reflected, specifically, by the extent of their (dis) agreement with these two statements, was associated with lower willingness to participate in DLC, reducing their likelihood of subscribing by about 11 percentage points (p < 0.001). More precisely, respondents who absolutely agreed that they fully trusted the utility displayed participation rates around 73%, whereas the rate dropped down to about 62% for those reporting utter distrust across these two closed-ended items. While this participation gap is still marked and noteworthy, it is far less than the 45 percentage point difference in participation that was found between respondents who did and did not volunteer (in response to the open-ended questions) distrust of the utility as a reason underlying their decision.

3.2. Impact of trust-restoring message on willingness to participate

We turn then to that second component of our study that went so far as to experimentally manipulate these feelings of trust and reassurance (vs. simply observing their 'natural' occurrence) to gauge the influence this might potentially exercise over willingness to participate in DLC. This is an important real-world problem for energy utilities, who around the globe often operate in complex political/regulatory environments and challenging market conditions, where gaining and keeping consumer trust is essential but hard to secure. Efforts to do so will often involve intensive (and expensive) marketing campaigns and community outreach efforts focused on (re)building brand image and winning (back) customer loyalty.

To test our second hypothesis, then, we analysed whether con-

19 The only factor that appeared to be as powerfully associated with unwillingness to participate was general disinterest and indifference (e.g., seeing no need for the program). For full details, see [92].

Table 1

Likelihood of participating in the DLC program as a function of (i) self-professed distrust of the utility and (ii) random assignment to a trust-restoring message.

Reasons offered for unwillingness to participate in the DLC program

If distrust was not mentioned If distrust was mentioned as as a barrier a barrier

Apparent influence of distrust on likelihood of participating

Trust-restoring If not exposed to message 62%

message (control group)

If exposed to message 65%

(treatment group)

18% 21%

-45 ppt -45 ppt

Effect of trust-restoring message on likelihood of participating

+3 ppt

+3 ppt

Note. ppt = percentage points.

If not exposed to trust-restoring message (control group)

If exposed to trust-restoring message (treatment group)

If distrust was not mentioned as a barrier If distrust was mentioned as a barrier Reasons offered for unwillingness to participate in the DLC program

Fig. 1. Varying willingness to subscribe to the DLC program as a function of self-professed distrust and experimentally-manipulated efforts at trust restoration.

veying a so-called 'trust-restoring' message had any impact on respondents' willingness to subscribe to the DLC program. As outlined in Section 2.1, our message (randomly assigned to half the respondents) consisted of a simple, two-sentence statement that explicitly acknowledged public disaffection with the utility, and provided reassurance of the utility's efforts to rebuild consumer confidence. As indicated earlier, our preliminary analyses revealed that the average likelihood of subscribing to the program was about 63% among eligible respondents. But we find that this level of participation in DLC, as predicted, did prove somewhat malleable: modestly responsive to our pointed trust-restoring message. We discerned a modest, but still noticeable difference in willingness to participate between those randomly assigned to the utility's 'reassuring' message (the 'experimental' group) and those not receiving any such reassurance (the randomly assigned 'control' group). Exposing respondents to this reassuring message generated a modest increase (although marginally significant at p = 0.073) in their willingness to subscribe to the DLC program, raising the likelihood of participation by around three percentage points (relative to the control group). Our results therefore suggest, in partial support of our second hypothesis, that if a utility explicitly acknowledges their loss of consumer trust, and reassures the public of their desire and efforts to rebuild it, this has the potential to at least modestly boost consumer willingness to participate in such a DLC program. We say ''at least" because in this case we did nothing more than insert just a couple of sentences - some bland, mildly worded assurances - into the text of a survey. We remain confident that a more engaging and widely promoted public campaign of this nature has even greater potential to boost con-

fidence in, and goodwill toward any utility that finds itself in this challenging position, but responds forcefully with some genuine trust-rebuilding endeavours.

3.3. Impact of trust-restoring message specifically on those expressing distrust

But will this kind of messaging campaign impact consumers differently, depending on their levels of self-professed distrust of the utility? Specifically, will reassuring messages of this nature have more influence on the initially distrustful, boosting willingness to participate in DLC to a greater degree for those coming to the campaign with low levels of trust at the outset? Again, these kinds of questions have important implications for utilities worldwide, bearing as they do upon the effective framing and targeting of such campaigns (and thus the efficient use of campaign expenditure and other marketing resources).

To test our final hypothesis, we examined whether our randomly assigned message would increase willingness to subscribe to the DLC program to a greater degree among those reporting distrust of the utility as a reason for not participating, than among those not raising any such concerns. To this point we have found that the detrimental influence of self-professed distrust of the utility far outweighed the modest impact of our (randomly assigned) trust-restoring message. We now find that (contrary to the expectations of our third hypothesis), exposure to our reassuring message had no greater impact in boosting willingness to participate, among those claiming their decision-making was influenced by

distrust of the utility, than among those expressing no misgivings of this kind.

Both of these findings are clearly illustrated in Fig. 1. Here we see that, among respondents not singling out distrust as a barrier to participation, likelihood of subscribing to the DLC program ranged from 62% (for the control group: receiving no message) up to 65% (for the treatment group: exposed to the trust-restoring message). But likewise, among respondents indicating distrust as a reason for their decision, likelihood of subscribing to the DLC program was similarly lifted 3 percentage points by the trust-restoring message, though from a substantially lower baseline to begin with, in this case rising from 18% (for the control group: receiving no message) up to 21% (for the treatment group: exposed to the trust-restoring message).

Again, two things are notable here. First, while starting from very different baselines, the modest improvements in respondents' likely participation seemingly wrought by the reassuring message - designed to rebuild trust - were essentially invariant (always an increase of about three percentage points), regardless of whether they had expressed any distrust of the utility. That is, there seemed to be no difference in the (either way, modest) impact of the trust-restoring message, as a function of distrust. This is readily apparent in Fig. 1, comparing either bar 1 against bar 2 (62% vs. 65%), or else, bar 3 against bar 4 (18% vs. 21%). In each case, the comparison reveals a 3 percentage point 'bump' attributable to the reassuring message, regardless of the starting stance of those being reassured. And second, in any case, this impact of our trust-restoring message continued to be dwarfed by the apparent influence of distrust on likely participation. The latter still diminished willingness to subscribe to the program by a weighty 45 percentage points or so, irrespective of any attempt to provide reassurance. Again, this is evident in Fig. 1 whether comparing bar 1 against bar 3 (62% vs. 18%), or bar 2 against bar 4 (65% vs. 21%).

4. Discussion

4.1. Summary of key findings

Reducing peak demand and balancing the supply of electricity on the network is a major priority of utilities and industry stakeholders worldwide. Demand management solutions such as direct load control devices represent a promising means of addressing this challenge. Yet to date, there has been relatively little robust scientific research conducted to understand and predict consumer response to such initiatives. Based on key insights from psychology and behavioural economics, we predicted that distrust may significantly impact consumer behaviour by serving as a decisionmaking shortcut or 'heuristic', particularly in situations characterised by high levels of choice, complexity, risk and uncertainty.

To investigate this notion further, we conducted a field-based survey-experiment examining the relationship between consumer distrust and self-reported willingness to subscribe to a DLC program to be offered by an Australian energy company. We hypothesised that respondents' self-professed distrust of the utility -volunteered in open-ended commentary explaining their decision - would be significantly associated with reduced willingness to register for the program (Hypothesis 1). Further, we proposed that exposure to a trust-restoring message - an explicit acknowledgment by the utility of pervasive consumer distrust, together with reassurances about their desire and efforts to rebuild community support - would significantly increase the probability of subscribing (Hypothesis 2). Finally, we hypothesised that this trust-restoring message would boost willingness to participate in the DLC program to a greater degree among those who singled out distrust in the utility as a reason for not participating (Hypothesis 3).

The first hypothesis was fully supported, the second partly so, and the third disconfirmed. In regard to the latter, exposure to the trust-restoring message had no greater (or lesser) impact on willingness to participate in the DLC program among those claiming to be influenced by distrust of the utility, than among those expressing no such misgivings. Of course, this could be either good or bad news for a utility finding itself in this position, depending on one's perspective, and one's willingness and capacity to respond appropriately. Our analysis suggests that well-considered campaigns to boost public trust and confidence in a utility are likely to be equally effective for both trusting and mistrustful consumers. And while the evidence (although we were able to test here only a rather mild reassurance, and a one-off message at that) seems to indicate that the returns to such 'rehabilitation' campaigns (and presumably similar strategies designed to build consumer confidence and restore reputation) might be modest, they at least appear unlikely to provoke a detrimental 'backlash' among sceptical segments of the public.

Although consumer distrust was evident among only a small segment of the sample, it nevertheless proved very intense and was powerfully associated with substantially lower willingness to participate in the DLC program. Moreover, while exposing consumers to a message designed to rebuild trust might improve participation rates, the magnitude of this improvement was, as noted, rather modest, and seemingly insufficient to mitigate the apparently substantial detrimental influence of distrust on willingness to participate. Thus, not only does distrust seem to play an important role in energy consumer decision-making, our results suggest that this influence may remain powerful even in the face of purpose-built and very pointed communications designed to restore public trust. Taking great pains to avoid losing consumer confidence in the first instance would seem to be far more efficient and effective a strategy.

4.2. Implications

The results from our study have important practical implications and highlight fruitful avenues for future research. In particular, our findings suggest that before introducing new technology such as energy demand management solutions, utilities should first seek to minimise and mitigate any distrust prevalent among consumers, while also making a concerted effort - taking proactive steps - to build and sustain a general sense of confidence and trust. In the presence of consumer distrust, resource-intensive advertising campaigns and marketing materials may fail to yield the return on investment that industry stakeholders hope for. They might even prove entirely ineffectual in motivating consumer uptake of such technology. Trust and distrust may serve as powerful decision-making tools when weighing the perceived costs and benefits of things that are novel and unfamiliar. This tends to be particularly pronounced when people confront complexity, risk and uncertainty, which are of course commonplace when choosing among new products and services. Consumers are bound to respond less positively when offered information and incentives by sources that they distrust, fear, or in any way find suspect. Consumer-focused messages to promote the uptake of demand management solutions should ideally be delivered by individuals and organisations that consumers perceive as competent, reliable, objective, fair, consistent and faithful - all attributes known to be highly conducive to trust [60,61,63,81].

One way to rebuild trust is by providing excellent customer service, both in terms of addressing immediate issues (reactive customer service) and anticipating future needs (proactive customer service). Our finding that a public acknowledgement and reassurance (somewhat akin to an apology) modestly improved willingness to participate, even among those who were distrustful,

confirms that an honest and straightforward approach to addressing past mistakes might go a considerable way toward rebuilding customer relationships. We concede that the impact of our randomly assigned trust-restoring message was dwarfed by the sheer weight of existing disaffection. But again, our reassuring message was very subtle, mild and brief: conveyed in just a few simple sentences unobtrusively embedded in the text of an unsolicited survey. This is vastly different to the kind of explicit, integrated and prolonged messaging typically conveyed in customer-focused marketing and advertising campaigns. As such, we are optimistic that a more targeted, purposeful media and community campaign to restore public confidence could have a much greater impact in reducing distrust, even to the extent of rendering consumers eager to participate in these kinds of DSM initiatives.

Another viable means of establishing and sustaining consumer confidence is to minimise the perceived risks, costs and losses associated with energy demand management. Much research has been undertaken to identify risk-reduction strategies ('risk relievers') that can markedly (and cost-effectively) influence consumer behaviour [94-98]. 'Relievers' of financial risks (e.g., offering discounts, rebates, lowest-price and money-back guarantees, no-cost returns/refunds, payment security) may be particularly useful for improving consumer acceptance of DSM initiatives, along with strategies to minimise perceived risks around time and effort (e.g., by making the purchase, installation and usage of the technology quick and easy). For example, offering consumers an obligationfree trial of the DLC program and/or a money-back guarantee may help to alleviate some of the perceived risks of participating, providing consumers with a 'safety net' - a sense of security and reassurance - around uncertainties and unknowns.

4.3. Potential limitations

While our study has yielded some important findings, it is not without potential limitations. First, although our field study was designed to be as realistic as possible, respondents still confronted a somewhat hypothetical scenario. While the DLC initiative described to them was certainly slated to be offered by the utility in the near future, ultimately our respondents were not actually signing up for the scheme. Rather, they were indicating their willingness to participate in a program that the utility was ''shortly going to be offering customers". Associated with this limitation was our necessary reliance not on objective measures of actual behaviour, but rather on respondents' self-reports, that is, on their subjective assessment of ''how likely [it was] that [they] would actually sign up for a program like the [name of DLC program]".

As noted earlier, we could not ask respondents to actually register for the program, as this was prohibited by both our research ethics committee and our third-party survey provider. Instead, we deployed what one might call a 'behavioroid' measure [93], assessing respondents' willingness to subscribe in the future (without having them actually do so). Given our ethical and practical constraints, this measure was thought to provide the closest analogue to the outcome of interest. That said, we do acknowledge that people's self-reports of their intentions and actions - past, present or future - are susceptible to well-known response biases and distortions, such as impression management and social desirability effects [99-103]. Self-reported behaviour may also be influenced by various aspects of data collection (tools and measures), including question order and wording, question context, and response formats [104]. Nevertheless, we designed our experiment and measures to be as realistic as possible, within the aforementioned practical and ethical constraints, and have gone to some lengths to minimise the error in our estimates.

We noted earlier that the rates of 'participation' observed among our sample (based on self-reported willingness to sub-

scribe) are likely to exceed the actual rates obtained in any real-world roll-out. In part this is because only about 20% of those approached for interview actually agreed to complete the survey. Bearing in mind that they were approached at large (rather than being drawn from some customer database), we can know nothing about the socio-demographic, attitudinal or behavioural attributes of those who declined even to respond to our survey, including their willingness or otherwise to sign up for direct load control. But it seems likely that compared with our survey participants, these non-respondents might be less trusting in general, less trusting of the utility in particular, and less willing to subscribe to its schemes.

We have already canvassed the likely impact of survey non-response in inflating apparent levels of enthusiasm for the load control scheme, while pointing out that our primary concern here remains not the level, but rather the determinants of uptake, especially the role played by consumer distrust in determining reluctance to subscribe. In regard to the latter, we expect that if our sample had also included this even-less-trustful 80% of consumers, it would have increased the profession of distrust in the sample, without substantially altering its apparent association with reluctance to subscribe.20 Moreover, recall that we found our trust-restoring message had the same impact (modestly improving willingness to sign on to the scheme) irrespective of anyone's professions of distrust. Accordingly, we expect that our assessment of the potential of trust-restoring communications to boost subscription to such a scheme is unlikely to be diminished by a broader, less amenable sample, or more disparate mass audience.

4.4. Directions for future research

Future research might usefully expand the breadth and depth of empirical studies in this domain, to better understand how consumers respond to diverse energy demand management initiatives for a range of domestic appliances - particularly those accounting for a disproportionate amount of energy consumption during peak periods. Curbing growth in peak demand is one critical means of placing downward pressure on rising electricity distribution costs [105]. Pinpointing precisely which energy-intensive appliances are the biggest contributors to peak demand, as well as those that consumers are most willing to place under load control, would be valuable for informing practical strategies to encourage uptake. For example, experimental evidence might indicate that consumers are far more accepting of DLC for pool pumps and hot-water systems than for air conditioners (or vice versa). These kinds of tailored insights can usefully inform the effective design and delivery of consumer-focused communications and other behavioural interventions, enhancing uptake of DLC on those specific appliances that consumers are most (and least) resistant to placing under load control.

Future research might additionally profit from exploring the efficacy of non-technological, behaviour-based strategies for motivating behaviour change in a way that reduces peak demand (for reviews of behavioural interventions to promote energy conservation and other pro-environmental behaviours, see [106-108]). While more rigorous experimental testing is clearly needed, strategies such as message framing [109-117]; the provision of simple attentional cues and prompts [118,119]; conveying information on both descriptive and injunctive social norms [120-124]; the use of goal-setting and commitment strategies [125-127]; providing clear, specific and immediate feedback [128-130]; and offering intrinsic and extrinsic rewards and incentives [131-133] are just

20 If anything (supposing it increased the variance in distrust within the sample), it might bolster the apparent impact of distrust on (un)willingness to subscribe.

some of the strategies that might prove effective. For example, tri-alling simple, low cost messaging that encourages householders to take shorter, cooler showers, to wash laundry in cold water and/or to lower the thermostats on their hot water systems, could potentially secure marked reductions in energy consumption associated with hot water usage. Messaging campaigns that might encourage consumers to shift from peak- to off-peak periods certain energy-intensive activities (e.g., showering, laundering, dishwashing) are also deserving of rigorous experimental testing. Carefully designed and tested customer-focused communications might prove useful for lowering peak usage of other energy-intensive appliances, e.g., messages about the social/health benefits of keeping cool on hot days by using alternative strategies other than conventional air conditioning [134], such as opening windows, sitting outside in the shade or a breeze, or going for a swim. Finally, no comprehensive assessment of DSM solutions would be complete without also testing consumer responses to more conventional financial incentives and 'price signals', particularly cost-reflective tariffs that charge consumers more for using electricity during peak periods. While experimental studies are accumulating [135-137], there is still vast scope to investigate the relative impact of different forms of dynamic pricing (e.g., time-of-use tariffs, peak time rebates, critical peak pricing, real-time pricing), particularly when 'bundled' together with other demand management initiatives like DLC.

To properly investigate consumer uptake (and usage) of these alternative demand management solutions, we require rigorous scientific studies that deploy randomised controlled experiments in real-world settings. Experimental designs and methods allow more precise and certain conclusions to be drawn regarding causal relationships between explanatory and outcome variables [33-35]. This is critically important for testing (and comparing) the true impact of any interventions - be they technological solutions such as DLC, or behaviour-driven strategies like message framing or financial incentives. Likewise, rigorous experimental testing will prove critical to determining the cost-effectiveness and scalability of alternative interventions for inducing the desired behaviour (i.e., measurable reductions in peak demand), both in absolute and relative terms.

Future experimental research should move beyond our study of what impact distrust can have on consumer choices, to examine when, where, how, why and for whom these effects occur. Attention should be focused on identifying the causal factors (predictors) and boundary conditions (moderators) that explain consumer responses to DSM initiatives, as well as the underlying processes and pathways (mediators) by which this might occur. A better grasp of these factors will greatly assist practitioners and policymakers in their efforts to design and deliver more cost-effective, scalable solutions that motivate optimal consumer behaviour and, ultimately, deliver network-wide benefits for the energy industry.

5. Conclusion

We conclude that trust and distrust serve as important decision-making 'heuristics' that consumers (whether consciously or not) will deploy when choosing whether to participate in automated demand management solutions like DLC. Energy utilities, practitioners and policymakers must remain mindful of the potential influence of trust and distrust when deciding how best to market and promote such solutions. This is especially true when a general sense of wariness, scepticism or suspicion is already extant among the public. We strongly encourage further rigorous research in this critical domain, particularly deploying the 'gold standard' of randomised controlled trials. The implications for theory, research and practice are far-reaching, and scientific rigour will be critical to

providing well-informed, evidence-based guidance for industry stakeholders around the globe.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/jj.apenergy.2016. 10.099.

References

EIA. Glossary. US Energy Information Administration, Independent Statistics and Analysis; 2016.

Palensky P, Dietrich D. Demand side management: demand response, intelligent energy systems, and smart loads. Ind Inform IEEE Trans 2011;7:381-8.

Li B, Shen J, Wang X, Jiang C. From controllable loads to generalized demand-side resources: a review on developments of demand-side resources. Renew Sustain Energy Rev 2016;53:936-44.

Pinson P, Madsen H. Benefits and challenges of electrical demand response: a critical review. Renew Sustain Energy Rev 2014;39:686-99. Brophy Haney A, Jamasb T, Platchkov L, Pollitt M. Demand-side management strategies and the residential sector: lessons from international experience. In: Jamasb T, Pollitt MG, editors. The future of electricity demand: customers, citizens and loads. Cambridge: Cambridge University Press; 2011. p. 337-78. Albadi MH, El-Saadany EF. A summary of demand response in electricity markets. Electric Power Syst Res 2008;78:1989-96.

Gellings CW. Saving energy and reducing CO2 emissions with

electricity. Lilburn, GA: The Fairmont Press, Inc.; 2011.

Aghaei J, Alizadeh M-I. Demand response in smart electricity grids equipped

with renewable energy sources: a review. Renew Sustain Energy Rev

2013;18:64-72.

FERC. 2010 Assessment of Demand Response and Advanced Metering: Staff Report. Washington, DC: Federal Energy Regulatory Commission; 2011. Available online: <https://www.ferc.gov/legal/staff-reports/2010-dr-report. pdf:>

Finn P, Fitzpatrick C. Demand side management of industrial electricity consumption: promoting the use of renewable energy through real-time pricing. Appl Energy 2014;113:11-21.

Kiliccote S, Olsen D, Sohn MD, Piette MA. Characterization of demand response in the commercial, industrial, and residential sectors in the United States. Wiley Interdiscip Rev: Energy Environ 2015. n/a-n/a. Samad T, Kiliccote S. Smart grid technologies and applications for the industrial sector. Comput Chem Eng 2012;47:76-84.

Yuemin D, Seung Ho H. A model of demand response energy management system in industrial facilities. In: Smart grid communications (SmartGridComm), 2013 IEEE international conference on. p. 241-6. Finn P, O'Connell M, Fitzpatrick C. Demand side management of a domestic dishwasher: wind energy gains, financial savings and peak-time load reduction. Appl Energy 2013;101:678-85.

Davito B, Tai H, Uhlaner R. The smart grid and the promise of demand-side management. McKinsey and Company; 2010.

Cappers P, Goldman C, Kathan D. Demand response in US electricity markets: empirical evidence. Energy 2010;35:1526-35.

Bhattacharyya SC. Energy demand management: concepts, issues, markets and governance. Energy Econ. London: Springer; 2011. p. 135-60. Newsham GR, Bowker BG. The effect of utility time-varying pricing and load control strategies on residential summer peak electricity use: a review. Energy Policy 2010;38:3289-96.

Ericson T. Direct load control of residential water heaters. Energy Policy 2009;37:3502-12.

Breukers S, Heiskanen E, Brohmann B, Mourik R, Feenstra C. Connecting research to practice to improve energy demand-side management (DSM). Energy 2011;36:2176-85.

Sovacool BK. What are we doing here? Analyzing fifteen years of energy scholarship and proposing a social science research agenda. Energy Res Soc Sci 2014;1:1-29.

Sovacool BK. Diversity: energy studies need social science. Nature 2014;511:529-30.

FERC. 2012 Assessment of Demand Response and Advanced Metering: Staff Report. Washington, DC: Federal Energy Regulatory Commission; 2012. Available online: <http://www.ferc.gov/legal/staff-reports/12-20-12-demand-response.pdf>.

ETSA_Utilities. Annual demand management compliance report. South Australia: ETSA Utilities; 2012.

Johnston MM. Direct load control and smart grid - customer issues for South Australia. Deakin West, ACT: St Vincent de Paul Society National Council; 2010.

NERA_Economic_Consulting. Cost benefit analysis of smart metering and direct load control: overview report for consultation Report for the Ministerial Council on Energy Smart Meter Working Group. Sydney, Australia: NERA Economic Consulting; 2008.

[27] Hedin M, Stimmel C. Demand response for residential markets: direct load control, time-of-use, critical peak pricing, and peak-time rebate programs for residential customers: global market analysis and forecasts. Boulder, CO, USA: Navigant Consulting, Inc.; 2012.

[28] Cook JD. Residential air conditioner direct load control ''Energy Partners Program". In: Ninth symposium on improving building systems in hot and humid climates. Arlington, TX.

[29] Lopes JS, Agnew P. FPL Residential Thermostat Load Control Pilot Project Evaluation. ACEEE Summer Study on Energy Efficiency in Buildings. Pacific Grove, CA 2010. Retrieved online 8 December 2014, from: <https://www. aceee.org/files/proceedings/2010/data/papers/1953.pdf>.

[30] KEMA, Comparison of California Investor-Owned-Utility (IOU) Direct Load Control (DLC) Programs. San Diego, CA. San Diego Gas and Electric; 2010. Retrieved online 8 December, 2014, from: <http://www.calmac.org/ publications/Final_report_for_California_DLC_Program_Comparison.pdf>

[31] Mecum S. A wish list for residential direct load control customers. ACEEE summer study on energy efficiency in buildings. Pacific Grove, CA: American Council for an Energy Efficient Economy (ACEEE). Retrieved online 8 December, 2014, from: <http://www.eceee.org/library/conference_ proceedings/ACEEE_buildings/2002/Panel_2/p2_15/paper>; 2002.

[32] Hackner R, Feldman S, LeClair J, Margossian B, Ozog M, Wallen D. How much load relief? From which households? Engineering residential direct load control in Wisconsin. ACEEE summer study on energy efficiency in buildings. Pacific Grove, CA: American Council for an Energy Efficient Economy. Retrieved 8 December, 2014, from <http://aceee.org/files/proceedings/ 1994/data/papers/SS94_Panel2_Paper14.pdf>; 1994.

[33] Vine E, Sullivan M, Lutzenhiser L, Blumstein C, Miller B. Experimentation and the evaluation of energy efficiency programs. Energ Effi 2014.

[34] Solomon P, Cavanaugh MM, Draine J. Randomized controlled trials: design and implementation for community-based psychosocial interventions. New York, US: Oxford University Press; 2009.

[35] Jadad AR, Enkin MW. Randomized controlled trials: questions, answers and musings. Chicester: Wiley-Blackwell; 2007.

[36] Rosenberger WF, Lachin JM. Randomization in clinical trials: theory and practice. New York: John Wiley & Sons; 2002.

[37] Schulz K, Altman D, Moher D. CONSORT 2010 Statement: updated guidelines for reporting parallel group randomized trials. Ann Intern Med 2010;152:726-32.

[38] Kendall JM. Designing a research project: randomised controlled trials and their principles. Emergency Med J 2003;20:164-8.

[39] Sibbald B, Roland M. Understanding controlled trials: Why are randomised controlled trials important? Br Med J 1998;316:201.

[40] Kwag H-G, Kim J-O. Reliability modeling of demand response considering uncertainty of customer behavior. Appl Energy 2014;122:24-33.

[41] Zuniga K, Castilla I, Aguilar R. Using fuzzy logic to model the behavior of residential electrical utility customers. Appl Energy 2014;115:384-93.

[42] Davis AL, Krishnamurti T, Fischhoff B, Bruine de Bruin W. Setting a standard for electricity pilot studies. Energy Policy 2013;62:401-9.

[43] Strengers Y. Beyond demand management: co-managing energy and water practices with Australian households. Policy Stud 2011;32:35-58.

[44] Saele H, Grande OS. Demand response from household customers: experiences from a pilot study in Norway. Smart Grid IEEE Trans 2011;2:102-9.

[45] Spence A, Demski C, Butler C, Parkhill K, Pidgeon N. Public perceptions of demand-side management and a smarter energy future. Nat Clim Change 2015;5:550-4.

[46] Morrison M, Kleinschafer J, Hicks J. Improving consumers' responsiveness to electricity demand management initiatives in regional New South Wales: the potential use of behavioural-based constructs for identifying market segments. Australasian J Regional Stud 2013;19:181.

[47] Fell MJ, Shipworth D, Huebner GM, Elwell CA. Public acceptability of domestic demand-side response in Great Britain: the role of automation and direct load control. Energy Res Soc Sci 2015;9:72-84.

[48] Frederiks ER, Stenner K, Hobman EV. Household energy use: applying behavioural economics to understand consumer decision-making and behaviour. Renew Sustain Energy Rev 2015;41:1385-94.

[49] Frederiks ER, Stenner K, Hobman EV. The socio-demographic and psychological predictors of residential energy consumption: a comprehensive review. Energies 2015;8:573-609.

[50] Stern P. What psychology knows about energy conservation. Am Psychol 1992;47:1224.

[51] Lutzenhiser L. Social and behavioral aspects of energy use. Annu Rev Energy Env 1993;18:247-89.

[52] Allcott H, Mullainathan S. Behavior and energy policy. Science 2010;327:1204-5.

[53] Carrico AR, Vandenbergh MP, Stern PC, Gardner GT, Dietz T, Gilligan JM. Energy and climate change: key lessons for implementing the behavioral wedge. J Energy Environ Law 2011;Winter 2011:61-7.

[54] Gigerenzer G, Gaissmaier W. Heuristic decision making. Annu Rev Psychol 2011;62:451-82.

[55] Kahneman D, Slovic P, Tversky A. Judgment under uncertainty: heuristics and biases. New York: Cambridge University Press; 1982.

[56] Tversky A, Kahneman D. Judgment under uncertainty: heuristics and biases. Science 1974;185:1124-31.

[57] Gilovich T, Griffin D, Kahneman D. Heuristics and biases: the psychology of intuitive judgment. Cambridge: Cambridge University Press; 2002.

[58] Kahneman D. Maps of bounded rationality: psychology for behavioral economics. Am Econ Rev 2003;93:1449-75.

[59] Hobman EV, Frederiks E, Stenner K, Meikle S. Uptake and usage of cost-reflective electricity pricing: insights from psychology and behavioural economics. Renew Sustain Energy Rev 2016;57:455-67.

[60] Renn O, Levine D. Credibility and trust in risk communication. In: Kasperson RE, Stallen PJM, editors. Communicating risks to the public. The Hague: Kluwer; 1991. p. 175-218.

[61] Earle TC, Cvetkovich GT. Social trust: toward a cosmopolitan society. Westport, CT: Praeger; 1995.

[62] Siegrist M, Cvetkovich G, Roth C. Salient value similarity, social trust, and risk/ benefit perception. Risk Anal 2000;20:353-62.

[63] Poortinga W, Pidgeon NF. Exploring the dimensionality of trust in risk regulation. Risk Anal 2003;23:961-72.

[64] Greenberg MR. Energy policy and research: the underappreciation of trust. Energy Res Soc Sci 2014;1:152-60.

[65] Craig S, McCann J. Assessing communication effects of energy conservation. J Consum Res 1978;3:82-8.

[66] Nader L, Milleron N. Dimensions of the ''people problem" in energy research and ''the" factual basis of dispersed energy futures. Energy 1979;4:953-67.

[67] Wustenhagen R, Wolsink M, Burer MJ. Social acceptance of renewable energy innovation: an introduction to the concept. Energy Policy 2007;35:2683-91.

[68] Huijts NMA, Molin EJE, Steg L. Psychological factors influencing sustainable energy technology acceptance: a review-based comprehensive framework. Renew Sustain Energy Rev 2012;16:525-31.

[69] Harrison McKnight D, Chervany N. Trust and distrust definitions: one bite at a time. In: Falcone R, Singh M, Tan Y-H, editors. Trust in cyber-societies. Berlin Heidelberg: Springer; 2001. p. 27-54.

[70] Van De Walle S, Six F. Trust and distrust as distinct concepts: why studying distrust in institutions is important. J Comparative Policy Anal: Res Pract 2014;16:158-74.

[71] Lewicki RJ, McAllister DJ, Bies RJ. Trust and distrust: new relationships and realities. Acad Manag Rev 1998;23:438-58.

[72] Rousseau DM, Sitkin SB, Burt RS, Camerer C. Not so different after all: a cross-discipline view of trust. Acad Manage Rev 1998;23:393-404.

[73] Govier T. Social trust and human communities. McGill-Queen's University Press; 1997.

[74] Clay-Warner J, Robinson DT. Social structure and emotion. Elsevier Science; 2011.

[75] Siegrist M, Cvetkovich G. Perception of hazards: the role of social trust and knowledge. Risk Anal 2000;20:713-9.

[76] Cofta P. Trust, complexity and control: confidence in a convergent world. Chichester, England: Wiley; 2007.

[77] Luhmann N. Trust and power. Chichester, England: Wiley; 1979.

[78] Cvetkovich GT, Lofstedt RE. Social trust and the management of trust. London: Earthscan; 1999.

[79] Terwel B. Origins and consequences of public trust: towards an understanding of public acceptance of carbon dioxide capture and storage. Netherlands: Leiden University; 2009.

[80] Midden CJ, Huijts N. The role oftrust in the affective evaluation ofnovel risks: the case of CO2 storage. Risk Anal 2009;29:743-51.

[81] Terwel B, Harnick F, Ellemers N, Daamen D. Competence-based and integrity-based trust as predictors of acceptance of Carbon Dioxide Capture and Storage (CCS). Risk Anal 2009;29:1129-40.

[82] Montijn-Dorgelo FNH, Midden CJH. The role of negative associations and trust in risk perception of new hydrogen systems. J Risk Res 2008;11:659-71.

[83] Siegrist M. The influence of trust and perceptions of risks and benefits on the acceptance of gene technology. Risk Anal 2000;20:195-203.

[84] Siegrist M. A causal model explaining the perception and acceptance of gene technology. J Appl Soc Psychol 1999;29:2093-106.

[85] Siegrist M, Cousin M, Kastenholz H, Wiek A. Public acceptance of nanotechnology foods and food packaging: the influence of affect and trust. Appetite 2007;49.

[86] Bronfman N, Lopez Vazquez E, Dorantes G. An empirical study for the direct and indirect links between trust in regulatory institutions and acceptability of hazards. Saf Sci 2009;47:686-92.

[87] Kasperson RE, Golding D, Tuler S. Social distrust as a factor in siting hazardous facilities and communicating risks. J Soc Issues 1992;48:161-87.

[88] Poortinga W, Pidgeon NF. Prior attitudes, salient value similarity, and dimensionality: toward an integrative model of trust in risk regulation. J Appl Soc Psychol 2006;36:1674-700.

[89] Terwel B, Harnick F, Ellemers N, Daamen D. How organizational motives and communications affect public trust in organizations: the case of carbon dioxide capture and storage. J Environ Psychol 2009;29:290-9.

[90] Terwel B, Harnick F, Ellemers N, Daamen D. Going beyond the properties of CO2 capture and storage (CCS) technology: how trust in stakeholders affects public acceptance of CCS. Int J Greenhouse Gas Control 2011;5:181-8.

[91] Costanzo M, Archer D, Aronson E, Pettigrew T. Energy conservation behavior: the difficult path from information to action. Am Psychol 1986;41:521-8.

[92] Frederiks E, Stenner K, Hobman EV. Why participate in energy demand management? An analysis of consumer motives. Unpublished manuscript; 2016.

[93] Aronson E, Ellsworth PC, Carlsmith JM, Gonzales MH. Methods of research in social psychology. New York: McGraw-Hill Publishing Company; 1990.

[94] Schiffman LG, Kanuk LL Consumer behavior. 10th ed. Englewood Cliffs, NJ: Prentice-Hall; 2009.

[99 100

101 102

110 111 112

Cases A-S. Perceived risk and risk-reduction strategies in Internet shopping. Int Rev Retail Distrib Consum Res 2002;12:375-94. Derbaix C. Perceived risk and risk relievers: an empirical investigation. J Econ Psychol 1983;3:19-38.

Roselius T. Consumer rankings of risk reduction methods. J Mark 1971:56-61. Mitchell V-W, McGoldrick PJ. Consumer's risk-reduction strategies: a review and synthesis. Int Rev RetailDistrib Consum Res 1996;6:1-33. Donaldson S, Grant-Vallone E. Understanding self-report bias in organizational behavior research. J Bus Psychol 2002;17:245-60. Moorman RH, Podsakoff PM. A meta-analytic review and empirical test of the potential confounding effects of social desirability response sets in organizational behaviour research. J Occup Org Psychol 1992;65:131-49. King MF, Bruner GC. Social desirability bias: a neglected aspect of validity testing. Psychol Mark 2000;17:79-103.

Paulhus DL. Socially desirable responding: the evolution of a construct. In: Braun HI, Jackson DN, Wiley DE, editors. The role of constructs in psychological and educational measurement. New Jersey: Lawrence Erlbaum Associates, Inc.; 2002. p. 51-76.

Holtgraves T. Social desirability and self-reports: testing models of socially

desirable responding. Pers Soc Psychol Bull 2004;30:161-72.

Schwarz N. Self-reports: how the questions shape the answers. Am Psychol

1999;54:93.

Graham PW, Brinsmead T, Hatfield-Dodds S. Australian retail electricity prices: Can we avoid repeating the rising trend of the past? Energy Policy 2015;86:456-69.

Osbaldiston R, Schott JP. Environmental sustainability and behavioral science: a meta-analysis of proenvironmental behavior experiments. Environ Behav 2012;44:257-99.

Schultz PW. Strategies for promoting proenvironmental behavior: lots of

tools but few instructions. Eur Psychol 2014;19:107-17.

Abrahamse W, Steg L, Vlek C, Rothengatter T. A review of intervention studies

aimed at household energy conservation. J Environ Psychol 2005;25:273-91.

Kahneman D, Tversky A. Choices, values and frames. Am Psychol

1984;39:341-50.

Tversky A, Kahneman D. The framing of decisions and the psychology of choice. Science 1981;211:453-8.

Cheng T, Woon DK, Lynes JK. The use of message framing in the promotion of environmentally sustainable behaviors. Soc Mark Quart 2011;17:48-62. Davis JJ. The effects of message framing on response to environmental communications. J Mass Commun Quart 1995;72:285-99. Ganzach Y, Karsahi N. Message framing and buying behavior: a field experiment. J Bus Res 1995;32:11-7.

Grewal D, Gotlieb J, Marmorstein H. The moderating effects of message framing and source credibility on the price-perceived risk relationship. J Consum Res 1994;21:145-53.

Kolandai-Matchett K. Mediated communication of'sustainable consumption' in the alternative media: a case study exploring a message framing strategy. Int J Consum Stud 2009;33:113-25.

Pelletier LG, Sharp E. Persuasive communication and proenvironmental behaviours: how message tailoring and message framing can improve the integration of behaviours through self-determined motivation. Can Psychol 2008;49:210-7.

Van de Velde L, Verbeke W, Popp M, Van Huylenbroeck G. The importance of message framing for providing information about sustainability and environmental aspects of energy. Energy Policy 2010;38:5541-9.

120 121

Thaler RH, Sunstein CR. Nudge: improving decisions about health, wealth,

and happiness. New Haven: Yale University Press; 2008.

Bekker MJ, Cumming TD, Osborne NK, Bruining AM, McClean JI, Leland LS.

Encouraging electricity savings in a university residential hall through a

combination of feedback, visual prompts, and incentives. J Appl Behav Anal

2010;43:327-31.

Allcott H. Social norms and energy conservation. J Pub Econ 2011;95:1082-95.

Ayres I, Raseman S, Shih A. Evidence from two large field experiments that peer comparison feedback can reduce residential energy usage. J Law Econ Organ 2012;29:992-1022.

Allcott H, Mullainathan S. Behavioral Science and Energy Policy. Cambridge, MA, USA: Working Paper, Massachusetts Institute of Technology (February); 2010.

Allcott H, Rogers T. The short-run and long-run effects of behavioral interventions: experimental evidence from energy conservation. Am Econ Rev 2014;104:3003-37.

Nolan JM, Schultz PW, Cialdini RB, Goldstein NJ, Griskevicius V. Normative social influence is underdetected. Pers Soc Psychol Bull 2008;34:913-23. Abrahamse W, Steg L, Vlek C, Rothengatter T. The effect of tailored information, goal setting, and tailored feedback on household energy use, energy-related behaviors, and behavioral antecedents. J Environ Psychol 2007;27:265-76.

Loock C, Staake T, Thiesse F. Motivating energy-efficient behavior with Green IS: an investigation of goal setting and the role of defaults. MIS Quart 2013;37:1313-32.

McCalley LT, Midden CJH. Energy conservation through product-integrated feedback: the roles of goal-setting and social orientation. J Econ Psychol 2002;23:589-603.

Midden CJH, Meter JE, Weenig MH, Zieverink HJA. Using feedback, reinforcement and information to reduce energy consumption in households: a field experiment. J Econ Psychol 1983;3:65-86. Seligman C, Darley JM. Feedback as a means of decreasing residential energy consumption. J Appl Psychol 1977;62:363-8.

Vine D, Buys L, Morris P. The effectiveness of energy feedback for conservation and peak demand: a literature review. Open J Energy Efficiency 2013;2:7-15.

Dolan P, Metcalfe R. Neighbors, knowledge and nuggets: two natural field experiments on the role of incentives on energy conservation. London School of Economics and University of Chicago; 2013.

Jacobsen GD, Kotchen MJ, Clendenning G. Community-based incentives for environmental protection: the case of green electricity. J Regul Econ 2013;44:30-52.

Mizobuchi K, Takeuchi K. The influences of financial and non-financial factors on energy-saving behaviour: a field experiment in Japan. Energy Policy 2013;63:775-87.

Nicholls L, Strengers Y. Air-conditioning and antibiotics: demand management insights from problematic health and household cooling practices. Energy Policy 2014;67:673-81.

Faruqui A, Sergici S, Akaba L. The impact of dynamic pricing on residential and small commercial and industrial usage: new experimental evidence from Connecticut. Energy J 2014;35.

Faruqui A, Sergici S. Household response to dynamic pricing of electricity: a survey of 15 experiments. J Regul Econ 2010;38:193-225. Faruqui A, Sergici S. Arcturus: international evidence on dynamic pricing. Electricity J 2013;26:55-65.