Sage Journals: Discover world-class research

Abstract

Based on a discrete choice experiment with 582 households in Switzerland, we find, that about 30% of the customers focus on price risks (cost focus) when they choose an electricity tariff, while 70% of the customers are more worried about volume risks (comfort focus). Customers with a cost focus, prefer contracts with low price risks and automatic load control, even when these contracts increased their volume risks and may lead to discomfort, while customers with a comfort focus are unlikely to choose a contract that exposes them to either price or volume risks.

All customers prefer direct load control of individual appliances to capacity subscriptions or other demand response approaches which limit their total electricity demand. While customers with a cost focus likely accept direct load control even if this reduces their comfort, enrolling customers with a comfort focus will require further efforts and contracts avoiding comfort loss.

Keywords

Residential electricity demand Consumer preferences Demand Response

1. Introduction

Sector coupling and increasing production from renewable energy sources are increasing the need for grid expansion and backup generation capacity. Demand response programs could reduce the need for such backup capacity by incentivizing customers to make their flexibility available to balance generation and demand and resolve grid congestions.

1.1 Types of Demand Response

Demand response is often grouped into incentive-based programs and price-based programs (Jordehi, 2019; Paterakis et al., 2017). Price-based programs, ask customers to adjust their load in response to price signals, such as static or dynamic time-of-use pricing, critical peak pricing, critical peak rebate, and real-time pricing. Incentive-based (or quantity rationing) demand response programs, on the other hand, ask consumers to “sell” their flexibility to third parties in return for compensation. While in the case of direct load control customers sell the right to restrict or curtail the load of specific appliances, restrictions in case of curtailable or interruptible service, capacity subscriptions (Doorman, 2005), and priority service (Wilson, 1997) apply to a household’s total electricity demand.

While there are many studies and field trials assessing the acceptance and performance of different price-based programs as well as direct load control (cf. section 1.3 below), very few studies and trials have compared the acceptance of incentive-based programs that restrict a customer’s total electricity demand to the acceptance of appliance-specific direct load control and price-based demand responses. Within this paper, we aim to fill this gap.

1.2 State of Demand Response in Switzerland

Historically, many distribution system operators in Switzerland have implemented direct load control of selected loads (e.g., warm-water boilers) using ripple control. Customers were not asked for consent or remunerated for their flexibility, as ripple control was operated in a manner that avoided comfort loss. Following the planned revision of the Swiss energy laws (Bundesrat, 2021), electric utilities now have to obtain explicit consent from customers for direct load control of their devices, or incentivize them to shift their loads, e.g., through time-of-use or capacity charges. In contrast to the U.S. or European countries, residential customers in Switzerland cannot choose their supplier and are obliged to purchase electricity from their local supplier.

To our knowledge, there are no publicly available statistics regarding customer enrolment in dynamic tariffs or explicit demand response programs in Switzerland. We have therefore assessed the state of demand response in Switzerland by verifying the tariff offerings of 26 large suppliers, which serve more than 60% of all residential customers in Switzerland.

We find that the tariff offered by most suppliers as default (16 of 26) was a uniform tariff with a constant price per kWh. The default tariff used by the remaining suppliers was a static time-of-use tariff with fixed high-tariff periods (usually during the daytime on weekdays). More than a third of the suppliers (10 of 26) offered optional tariff discounts for loads that can be directly controlled (such as heat pumps, boilers, or EV charging stations), usually in combination with a time-of-use tariff. None of the utilities offered a critical peak price, and only one utility offered an optional dynamic time-of-use price contract. While we do not know how many customers signed up for each tariff, international experience suggests that tariff uptake for optional tariffs is rather low (cf. next section). We therefore assume that most customers in Switzerland are either subscribed to a uniform tariff or a time-of-use tariff.

1.3 Assessments of Demand Response Participation

Many studies have assessed the acceptance of different price-based demand response programs (Buryk et al., 2015; Dutta & Mitra, 2017; Nicolson et al., 2018; Parrish et al., 2020). They consistently find that consumers prefer tariffs with lower price risks, that is a lower frequency and a lower magnitude of price spikes. However, the relative importance of these attributes compared to other contract characteristics is often not explored.

Studies regarding the acceptance of incentive-based demand response programs such as direct load control have found, that acceptance depends on factors such as trust (Fell, 2016; Goulden et al., 2014; Lopes et al., 2016; Stenner et al., 2017; Throndsen & Ryghaug, 2015), perceived control and in particular advance notice (Curtis et al., 2020) and the possibility to override direct load control (Buchanan et al., 2016; Curtis et al., 2020; Dütschke & Paetz, 2013; Fell et al., 2015; Xu et al., 2018) as well as the framing of the intervention (Becker et al., 2016; Burchell, 2016). While most of these papers have explored whether direct load control of specific appliances or devices is acceptable, there are no studies that have compared the acceptability of controlling individual loads to the acceptability of capacity subscriptions, which would result in a disconnection of the total load of consumers when they exceed the subscribed capacity during scarcity situations. Within this study we plan to fill this gap.

Studies assessing actual customer enrolment find, that stated tariff acceptance is about 5 times higher than actual tariff uptake (Nicolson et al., 2018). While trials and commercial programs using opt-in recruitment typically achieved enrolment rates between 5 and 50%, opt-out studies tend to achieve an enrolment of more than 70% of customers (Nicolson et al., 2018; Parrish et al., 2019; Potter et al., 2014). Compared to the impact of opt-in versus opt-out recruitment, the impact of other factors, such as the type of demand response program is much less significant, leading some authors to conclude that enrolment for different pricing plans is significant but small (Potter et al., 2014), while others conclude that there is “no obvious pattern in rates of recruitment across different types of demand response” (Parrish et al., 2019).

Regarding the acceptance of priority service, Hayn found that more than 75% of respondents stated that they would be willing to accept a capacity subscription (Hayn, 2016). However, the survey did not compare this to the acceptance of other contract types, which are associated with a lower volume risk. Using focus groups, Darby and Pisica found that the acceptance of capacity subscriptions was much lower than for time-of-use pricing, in a similar range as that of critical peak pricing and real-time pricing (Darby & Pisica, 2013). However, due to the small sample size, they neither tested the statistical significance of their results nor quantified the relative importance of the volume risks resulting from capacity subscriptions or direct load control of individual appliances to the price risks resulting from contracts with dynamic prices.

Design preferences for direct load control programs have also been assessed by (Yilmaz et al., 2021, 2022), using the same method as in this paper, i.e. based on discrete choice experiments and latent class analysis. However, they focused on different attributes, such as the duration of the intervention, override option and advance notice, and did not compare the acceptability of timevarying prices and capacity subscriptions to direct load control.

1.4 Aim of this study

The purpose of this study is to fill the above-mentioned gaps and assess the:

relative acceptability of volume risks from capacity subscriptions (limiting total demand) compared to volume risks from direct load control (limiting individual appliances) and price risks from time-varying price contracts,

drivers of customer heterogeneity regarding risk preferences,

economic efficiency of load control approaches with different levels of price and volume risk.

1.5 Structure of the paper

The rest of this paper is structured as follows. Section 2 describes the method for survey design, data collection, and analysis of responses that we have used, section 3 presents the empirical results, section 4 discusses implications regarding the trade-off between grid expansion and demand response and section 5 concludes.

2. Method

To answer our research questions, we designed an online survey including a discrete choice experiment. A choice experiment is a stated preference method that is widely used in many disciplines to elicit the values of private and public goods, with the estimates potentially suffering from hypothetical bias (Neill et al., 1994) and strategic bias (Whittington et al., 1990) as well as other sources of bias, such as loss aversion (Kahneman et al., 1991), scope sensitivity (Dugstad et al., 2021), and the construction of preferences (Schkade & Payne, 1994). Through our survey design, data collection, and responses evaluation, we aim to mitigate these biases following the contemporary guidance for stated preference studies (Johnston et al., 2017).

2.1 Survey Design

An overview of the main survey sections is provided in Figure 1.

Figure 1:

Main sections of the survey.

Following an introduction, participants were asked to complete a discrete choice experiment regarding tariff choice. Discrete choice experiments have already been used by several other studies regarding tariff choices (Lehmann et al., 2021; Yilmaz et al., 2019). Compared to other elicitation approaches, the main advantage of this method is its replication of a real-life choice situation, where consumers also must choose between products that are characterized by a bundle of attributes rather than a separate specification of attribute importance, as in the case of other elicitation methods. The analysis of results is based on random utility and will be further explained below.

To mitigate the potential hypothetical bias arising from a choice experiment using stated preferences, in the introduction section we asked participants about their own practical experience with different tariff types (cf. Appendix A, Figure 12). In addition to that, at the end of the choice experiment we ask participants to report whether they truly understand the choice tasks through follow-up questions.

For the tariff choice experiment, participants were randomly assigned to 1 of 14 blocks containing 6 choice tasks. In each of these tasks, participants were asked to select their preferred option from 3 tariffs that differed regarding the peak frequency, additional costs during peak periods, savings target, automatic action, and monthly base fee. For each of these attributes, we defined three different levels, which are shown in the sample choice task in Figure 2. As recommended by Goldar and Misra, the monetary values which we calculate are therefore based on customers’ willingness to pay, rather than their willingness to accept compensation (Goldar & Misra, 2001), which to some extent could also help to reduce the hypothetical bias. In addition, the six choice tasks were presented to participants in random order to mitigate the start-point bias.

Figure 2:

Example choice task.

The experiment was designed using Sawtooth’s “complete enumeration” function, which ensures that attribute levels are statistically independent and that two-way frequencies of level combinations between attributes are balanced (Chrzan & Orme, 2000).

To ensure that participants understood the survey, the attributes of the discrete choice experiment were first described followed by a reading example (cf. Appendix A below) before participants were asked to complete the choice task. In addition to that, during the choice task they received the mouseover information described on the right side of Figure 2 when they hovered over the respective items.

Following the tariff choice experiment, consumers were asked to complete several follow-up questions to help us analyze how they constructed their preferences (Schkade & Payne, 1994). This included questions regarding survey understanding, their response strategy, and their risk aversion (cf Figure 3).

Figure 3:

Follow-up questions regarding survey understanding (top), response strategy (middle) and risk-aversion (bottom).

Additional data regarding demographics, respondents’ current electricity contract, and equipment were obtained as part of the SHEDS survey from Intervista (see next section).

2.2 Data collection

To reduce sample selection bias (McFadden, 2017), surveys were carried out as part of the Swiss Household Energy Demand Survey (SHEDS) (Weber et al., 2017), using a representative online panel of incentivized respondents from Intervista, a professional Swiss survey agency.

The initial sample for our analysis consisted of 582 respondents, who had fully answered the survey. A short comparison of the demographics from this sample and Switzerland is shown in Table 1. Apart from the classes that are marked with a footnote, the sample is reasonably representative.

Table 1:

Comparison of demographics for the sample and Switzerland as a whole.

Variables	Value	Switzerland	Sample
Region	German Kanton	70.7%	75.9%
	French Kanton	24.9%	24.1%
	Italian Kanton^a	4.4%	0.0%
Gender	Male	49.6%	51.5%
	Female	50.4%	48.5%
Age	[0.0, 10.0)^a	9.1%	0.0%
	[10.0, 20.0)^a	9.8%	0.0%
	[20.0, 30.0)	12.2%	9.5%
	[30.0, 40.0)	14.2%	22.3%
	[40.0, 50.0)	14.2%	18.6%
	[50.0, 60.0)	15.2%	23.9%
	[60.0, 70.0)	11.1%	14.3%
	[70.0, 80.0)	8.5%	10.1%
	[80.0, 90.0)^a	4.6%	1.4%
	[90.0, 99.0)^a	1.2%	0.0%
Income	Less than 3,000^a	7.1%	4.3%
	3,000–4,500	10.1%	9.8%
	4,501–6,000	12.2%	14.8%
	6,001–9,000	25.0%	22.9%
	9,001–12,000	19.9%	19.8%
	More than 12,000^a	22.3%	11.9%
Education	Compulsory school^a	14.6%	1.7%
	Secondary school	40.0%	51.4%
	Tertiary education	27.3%	46.9%
Employment	Employed	50.2%	68.6%
	Unemployed	2.4%	2.6%
	Retired	17.8%	18.7%
	Other	11.5%	10.1%
Household size	1 Person	17.1%	27.0%
	2 Persons	27.5%	43.6%
	3 Persons	14.2%	11.3%
	4 Persons^a	20.9%	12.7%
	5 Persons, or more^a	10.0%	5.3%
Living situation	Tenant	56.9%	58.6%
	Owner	43.1%	41.4%

Notes: ^a The sample contains less than 70% of the respondents required for a representative sample.

2.3 Analysis of Responses

The analysis of discrete choice experiments is based on random utility theory, which assumes, that the utility of an alternative can be described as the sum of the utilities for each of its attributes, and that consumers choose the best alternative which gives the highest utility (Louviere et al., 2000; Train, 2009). An overview of our approach for analysis is provided in Figure 4.

Figure 4:

Main steps of analysis.

To detect scope sensitivity and avoid imposing a functional form on respondent preferences, we use dummy encoding for all attribute levels (Dugstad et al., 2021).

In step 1, we analyzed the responses using a simple conditional logit model, which assumes that the utility of each attribute and level is the same across all respondents, and that random error terms are identically and independently distributed. The probability that respondent n chooses an alternative j in the choice task t can be described as:

$P_{njt} (β) = \frac{\exp (β^{'} X_{njt})}{\sum_{j = 1}^{J} \exp (β^{'} X_{njt})}$ (1)

where β is the vector of utilities for each attribute, and X_njt is the vector of attribute levels of alternative j in choice task t of respondent n. We calibrate the utilities using the Logit function in Lighthouse Studio 9.8.0 (Sawtooth Software Inc, 2021), which fits McFadden’s conditional logit model (McFadden, 1973).

In step 2, we analyze the heterogeneity of preferences using a latent class conditional logit (LCL) model (Greene & Hensher, 2003), which assumes, that the utility of each attribute and level depends on the “class” to which a respondent belongs. In contrast to a mixed logit model with subsequent respondent clustering, e.g., using k-means, the latent class approach does not require us to make any assumptions about the distribution of parameters across individuals (Greene & Hensher, 2003) and provides us with class membership probabilities, that can be used to quantify the uncertainty regarding class membership (Sinha et al., 2021; Weller et al., 2020). If respondent n belongs to class c, the probability that the respondent chooses alternative j in choice task t can be described as follows in a latent class model:

$P_{njt} (β_{c}) = \frac{\exp (β_{c}^{'} X_{njt})}{\sum_{j = 1}^{J} \exp (β_{c}^{'} X_{njt})}$ (2)

Where β_c is the vector of the class-specific utilities of class c for each attribute level, and X_njt is the vector of attribute levels of alternative j in choice task t of respondent n. We calibrate the utilities using the latent class function in Lighthouse Studio 9.8.0 (Sawtooth Software Inc, 2021) assuming between 1 and 7 classes. As there is no single best method or criterion to determine the optimal number of classes, we compare various indicators including the Bayesian Information Criterion, the Consistent Akaike Information Criterion, the Bayes Factor, the correct model Probability and the average posterior class probability across the models with different numbers of classes (Little, 2013; Nylund-Gibson & Choi, 2018; Sinha et al., 2021; Weller et al., 2020).

In step 3, we analyzed to what extent the probability π_n,c, that respondent n belongs to class c, which we have obtained in step 2, depends on socio-demographic variables including age, gender education, and respondents’ attitude and response strategy through an ordinary least squares regression of π_n,c, as a function of these variables.

In step 4, we finally performed several consistency checks, to make sure that our results were not influenced by random responses.

3. Results

3.1 Uniform Utilities

The calibrated relative utilities of the conditional logit model for each attribute level are displayed in Figure 5.

Figure 5:

Utilities of the conditional logit model.

Overall, the utilities show the expected trend: a higher peak frequency, a higher peak cost, a more restrictive savings target, more restrictive automatic actions, or a higher monthly fee eventually all lead to a lower utility. The attribute that has the highest impact on utility is the monthly fee, with an impact that is more than twice as large as that of other attributes. A stronger savings target (=higher volume risk) creates a slightly larger disutility than an increasing peak cost (=higher price risk). The disutility increases further if automatic actions restrict total demand rather than individual appliances, which indicates a higher acceptance of direct load control compared to capacity subscriptions. Another striking point is that the utility of the “middle” attribute level is higher than the average utility of the “high” and the “low” attribute levels, even in case of a linearly increasing or decreasing attribute levels, such as monthly fees or savings target. In case of the savings target, the utility of the “middle” value of “maximum 1 big appliance” is even higher than that of a higher or lower savings target. As we will see in the subsequent section, this can be partly explained by the heterogeneity of preferences for different user classes. Compared to the null model, our model using uniform utilities improves the log likelihood by about 10%.

3.2 Heterogenous Utilities

To estimate the optimal number of latent classes, we fit a latent class model with 1 to 7 classes and compare the resulting values of several information criteria (Figure 6, a), relative fit statistics and classification diagnostics (Figure 6, b) (Little, 2013; Nylund-Gibson & Choi, 2018; Sinha et al., 2021; Weller et al., 2020).

Figure 6:

Information criteria (a), and relative fit statistics and diagnostics (b) for different numbers of latent classes.

Regarding the information criteria, the elbow point of the Bayesian Information Criterion (BIC) and the Consistent Akaike Information Criterion (CAIC) suggest the usage of 2 latent classes, while the absolute minimum value of the CAIC occurs for 3 classes, and for the BIC in case of 4 classes. Regarding the relative fit statistics, the Bayes Factor is highest for models with 4 to 6 classes, but remains below the threshold value of 3, which indicates rather weak evidence (Little, 2013; Nylund-Gibson & Choi, 2018). Likewise, the correct model probability is highest in case of 3–4 classes but would also allow the usage of between 2 to 5 latent classes. The classification diagnostic of the average posterior class membership probability finally continues to drop as the number of classes increases, but always remains above 0.8, indicating a relatively clear separation of the classes in all cases.

Taken together, the different indicators seem to support a selection of between 2 to 4 latent classes. In the remainder of this paper, we present the results for a 2 class solution, as this is the most parsimonious model supported by the indicators (Little, 2013, S. 571). The relative size and preference of the customer segments we identified in case of 2 classes remains similar in case of a model with 3 or 4 classes, so that the overall message would not change, while the addition of 1 or 2 smaller classes increases the risk of over-interpretation. (Solutions for the 3 and 4 class models were presented to the reviewers.)

The resulting utilities for the latent class analysis with 2 respondent classes are shown in Figure 7. For most attributes, the relative utility of attribute levels follows a similar shape as in Figure 5, and regarding automatic action customers from both classes continue to prefer limitations of individual appliances (direct load control) to limitations of total demand (capacity subscriptions). However, with regards to the savings target, the two respondent classes seem to have opposite preferences.

Figure 7:

Utilities of the latent class model assuming two classes.

Respondents who belong to class 1 with a cost focus (about 30% of respondents), would prefer a stronger savings target resulting in a higher volume risk and potentially discomfort— presumably to avoid peak cost. This is also supported by their much stronger aversion to higher monthly fees and higher peak cost.

By contrast, respondents who belong to class 2 with a comfort focus (about 70% of respondents) would prefer a weaker savings target, exposing them to a lower volume risk and less discomfort. Respondents from this group are also worried about price risks in terms of a higher peak frequency or higher peak cost. Yet, a reduction of their volume risks is more important to them than a reduction of their peak frequency or their peak cost. In addition, customers from class 2 tend to react to extreme values in each category as the utility difference between the first two attribute-levels is always smaller than the change in utility when moving to the third attribute value. Some of the respondents had even explicitly stated, that they tried to choose “not the least expensive tariff, but an intermediate version” (cf. Response strategy: achieve threshold in Table 2 below).

Table 2:

OLS regression coefficients of respondent characteristics on class membership probabilities

Variables	Membership probability for:
	Class (1)	Class (2)
Age	0.00	0.00
Education: secondary school	–0.13	0.13
Education: tertiary education	–0.09	0.09
Employment: other	–0.02	0.02
Employment: retired	–0.06	0.06
Employment: unemployed	–0.171*	0.171*
Gender: male	0.03	–0.03
Household size	–0.034**	0.034**
Income: 3,000–4,500	–0.04	0.04
Income: 4,501–6,000	0.01	–0.01
Income: 6,001–9,000	0.01	–0.01
Income: 9,001–12,000	–0.06	0.06
Income: more than 12,000	–0.109*	0.109*
Living situation: tenant	–0.02	0.02
Region: German kanton	–0.02	0.02
Respondent contract: automatic load control	–0.02	0.02
Respondent contract: default tariff	–0.063*	0.063*
Respondent contract: green tariff	–0.093*	0.093*
Respondent contract: variable price	–0.02	0.02
Respondent equipment: electric vehicles	–0.07	0.07
Respondent equipment: smart-meter	–0.21	0.21
Respondent equipment: heat pump	0.00	0.00
Response strategy: 1 large device	0.12	–0.12
Response strategy: achieve threshold	–0.04	0.04
Response strategy: comfort focus	–0.125***	0.125***
Response strategy: cost focus	0.256***	–0.256***
Response strategy: min peak frequency	–0.09	0.09
Response strategy: random response	0.06	–0.06
Response strategy: risk-taking	0.00	0.00
Survey: part of treatment group	–0.04	0.04
Survey: understanding of choice-tasks	0.05	–0.05
constant	0.445***	0.555***
Observations	582	582
Adjusted R2	0.13	0.13
Residual Std. Error	0.37	0.37

Notes: The coefficients from the ordinary least squares regression are presented in the table. *p<0.1; **p<0.05; ***p<0.01

To understand potential drivers of these preferences, we will regress the membership probabilities for each of these classes on several demographic and psychographic variables in the following section.

3.3 Drivers of Heterogeneity

Table 2 shows the results of regressing the respondents’ probability of belonging to each of the latent classes on several demographic and psychographic variables. Response strategies have been coded based on the free-text answers that were provided by the participants. It turns out that the most important driver of class membership, is respondents’ stated response strategy. Respondents who said they were calculating the least costly tariff, or selecting the tariff based on the price per month or per kWh (= “Response strategy: cost focus”) were 26% more likely to belong to class 1. On the other hand, respondents, who said they were trying to avoid too restrictive savings targets or automatic actions (= “Response strategy: comfort focus”) were 12.5% more likely to belong to class 2.

With regards to the demographic variables, on average an increasing household size was associated with a 3% higher probability of belonging to class 2 (comfort focus) and a 3% lower probability of belonging to class 1 (cost focus) for an additional member of the household.

If we include demographics that were only relevant at the 10% significance level, it seems that the probability of having a comfort focus increases for unemployed respondents (a 17% higher probability) or respondents with a monthly income above 12’000 CHF (an 11% higher probability). In addition, respondents with a comfort focus are more likely to remain on their default tariff (a 6% higher probability) or choose a green tariff (a 9% higher probability).

3.4 Consistency checks

Despite our efforts to explain survey attributes (cf. section 2.1 and Appendix A), we cannot rule out that at least some of the respondents did not understand the survey. To verify whether our findings were distorted by incomplete understanding, we performed two consistency checks.

As a first step, we conducted a pre-test with 16 respondents before launching the main survey. Most pre-test respondents indicated that they had either “Completely understood” (4 respondents), or “Somewhat understood” (11 respondents) the choice task. Only 1 respondent indicated that he “Didn’t understand” the choice task. We therefore used a very similar version of the choice task for the main survey.

As a second step we identified respondents to the main survey, that seemed to have a random response strategy, by calculating the root-likelihood of the survey responses for each respondent based on a Hierarchical-Bayes Model (Johnson, 2000). In Figure 8, we then compared to what extent the responses to the filtering questions in Figure 3 (top, and middle) were associated with a lower root-likelihood.

Figure 8:

Impact of survey understanding (a) and response strategy (b) on respondent root- likelihood (RLH).

As we can see, respondents who said that they “Didn’t understand” the choice task (cf. Figure 8, a) or felt their response strategy was a “Random choice” (cf. Figure 8, b) achieved a slightly lower median root-likelihood. However, the effect is relatively small. Most of these respondents still achieved a root-likelihood above 0.6, while in case of a fully random choice, we would have expected a root-likelihood of 0.3. Instead of discarding these responses from the start, we have thus reported the results including all respondents in the previous sections.

As a third step, we tested to what extent the conclusions from the main survey regarding the attribute utilities in section 3.2 would be different, if we discarded:

Respondents based on the filtering questions: removing respondents who said they “Didn’t understand” the survey, or had used a “Random choice” strategy;

Respondents based on filtering questions and root-likelihood: additionally removing respondents with a root-likelihood below 0.6 from the remaining dataset.

In Figure 9, the optimal number of classes is the same for both filtered datasets. For comparison purposes over different sample sizes, the original Bayesian Information Criterion has been normalized to a maximum value of 1.

Figure 9:

Impact of different respondent filters on the Bayesian Information Criterion.

In Figure 10, the part-worth utilities for each of the classes are almost unchanged if the filters are applied. Focusing on the results of the whole dataset will thus not distort the conclusions.

Figure 10:

Impact of different respondent filters on the part-worth utilities of group 1) Cost focus (a) and group 2) Comfort focus (b).

4. Stylized Net Benefit of Different Tariffs

Within this section we discuss the results of a stylized analysis regarding the economic benefit of switching from a flat tariff¹ to a time-of-use tariff, critical peak prices, direct load control of individual appliances or a capacity subscription. Attribute levels were chosen to resemble real world tariffs as closely as possible. However, when a real-world attribute level (such as peak frequencies exceeding 100 days per year, or an additional peak-cost of less than 1 CHF/peak) had not been tested in the survey, we chose the closest matching attribute level from the survey (e.g. “Peak-Frequency” = “100 days” and “Peak-Cost” = “CHF 1”). Actual costs and benefits may differ from these assumptions. In addition, survey answers are only a weak indication of actual tariff uptake, as tariff switching is associated with additional hassle, and tariffs may be framed in a more positive (non-disruptive) way in reality. The results of the following cost-benefit analysis should thus be interpreted with caution.

We first calculated the customer benefit of switching from a flat-price contract to other contracts (cf. Appendix B, section 1). The resulting customer benefits are displayed as blue bars in Figure 11. As we can see, customers from class 1 (Figure 11 a) prefer a contract with direct load control or capacity subscriptions and would be willing to pay about 2 CHF per month for a switch to capacity subscriptions and about 3 CHF per month for a switch to direct load control of individual appliances. Customers from class 2 (Figure 11 b) prefer a contract with flat prices and would need to be paid 17–19 CHF per month to accept switching to time-of-use pricing or critical peak prices and considerably more to accept direct load control or capacity subscriptions.

Figure 11:

Customer benefit and net benefit of switching from a flat price contract to other approaches for group 1) Cost focus (a) and group 2) Comfort focus (b).

We then calculate the net benefit as the sum of customer benefits and grid benefits, minus implementation costs (cf. Appendix B, section 2 and 3). The resulting net benefits are displayed as orange bars in Figure 11. For critical peak prices and direct load control, net benefits are very similar to the customer benefits because our estimate of grid benefits is similar to our estimate of implementation costs. For time-of-use tariffs and capacity subscriptions, our (rough) estimate of implementation costs exceeds grid benefits. In case of capacity subscriptions this is due to the low estimate of grid-benefits in (Hayn, 2016). We expect studies including heat-pumps and electric vehicles to result in higher grid benefits, more similar to those achieved by direct load control.

However, while net benefits of customers from class 1 (Figure 11 a) are so small (about −6 to +4 CHF per month) that different assumptions regarding grid benefits and/or implementation costs could change the direction of net benefits (e.g., result in a positive net benefit for time-of-use tariffs and negative net benefit for capacity subscriptions), net benefits of customers with a comfort focus (Figure 11 b) are so strongly negative (less than −17 CHF per month) that different assumptions regarding grid benefits and implementation costs are unlikely to result in a positive net benefit for any of the contract types. Recruiting customers from this group for demand response will thus require additional measures to improve customer benefits (see (Parrish et al., 2020)).

5. Conclusions

Our analysis shows that some residential electricity consumers are more averse to price risks (to avoid high costs), while others care more about volume risks (to avoid discomfort).

In our survey sample, about 30% of the customers were more concerned about price (cost focus) than volume risks. As a result, these customers, who were more likely to live in smaller households, preferred a contract with a strong energy savings target and automatic load control to a flat price contract without load control as means to reduce their energy bills, even if this could reduce their comfort.

Most customers in our sample, more than 50%, cared more about volume risks (comfort focus) than about price risks. As they were also averse to price risks, customers from this group, which were more likely to live in larger households, were unlikely to sign up for any contract that could reduce their comfort by exposing them to automatic load control or higher peak-prices, even if this could reduce their monthly bill.

Regarding the type of automatic load control, customers from both groups seemed to prefer a direct load control of individual appliances to capacity subscriptions or other demand response approaches affecting their total load. A stylized analysis regarding the net benefits of different load-control approaches points in the same direction, as the studies we have found in literature seem to suggest stronger peak-load reductions through direct load control than through capacity subscriptions at similar implementation costs.

While our survey was based on customers’ stated preferences, meta-analyses of customer enrolment in demand-response programs suggest similar participation rates with a range of 10% to 50% in the case of opt-in recruitment (Nicolson et al., 2018; Parrish et al., 2019; Potter et al., 2014). Apart from customers’ preference for different tariff schemes and load-control approaches which we assessed in this study, the actual tariff switching may be influenced more strongly by the recruitment strategy (opt-in vs. opt-out) and other non-price factors (Deller et al., 2021; He & Reiner, 2017).

For countries such as Switzerland, which are planning to roll out smart meters and expect the connection of many new loads (such as electric vehicles and heat pumps) (SFOE & Prognos, 2020), it could therefore make sense to install smart meters and flexible loads which are automated without comfort loss as a default (opt-out recruitment) or at least “automation ready”, as long as the additional cost for the load control functionality is small. Accompanying the roll-out by an information campaign regarding the benefits of automation (such as potential comfort increases, bill reductions, or environmental benefits as a result of smart automatic control) (Parrish et al., 2020) could increase customer enrolment in demand response, which would reduce the cost and need for grid expansion (Consentec et al., 2022). In addition to bill guarantees (Ludwig & Winzer, 2022), notifications and override options (Yilmaz et al., 2022) our survey results illustrate the importance of allowing customers to choose between direct load control contracts with different comfort levels, as about 30% of the customers may prefer a contract which exposes them to comfort losses to reduce their electricity bill. At the same time, an “automation readiness” of all newly installed smart-meters, heat-pumps and electric vehicle charging stations would enable faster switching to automatic load control, where this is preferred because of changing consumer preferences (e.g., new tenants moving in), or as an emergency measure during future energy crises (e.g., as last resort measure to avoid rotating black-outs).

Despite our efforts to mitigate response biases, our estimates of consumers’ willingness to pay to avoid price and volume risks may still suffer from hypothetical and strategic biases. Additional methods could be used in future work separately or in combination to address these biases, such as consequentiality of choices (Vossler et al., 2012), inferred valuation (Lusk & Norwood, 2009), honesty priming (Howard et al., 2017), and budget constraint reminding. In addition, studies that compare our stated preference estimates with revealed preference ones from randomized field experiments, when such actual experiments exist, could significantly improve the design of demand response programs. In particular for capacity subscriptions there is still a lack of empirical data that quantifies grid benefits and implementation costs compared to other demand-response approaches.

Supplemental Material

sj-pdf-1-enj-10.5547_01956574.45.2.cwin – Supplemental material for Cost Focus versus Comfort Focus: Evidence from a Discrete Choice Experiment with Swiss Residential Electricity Customers

Supplemental material, sj-pdf-1-enj-10.5547_01956574.45.2.cwin for Cost Focus versus Comfort Focus: Evidence from a Discrete Choice Experiment with Swiss Residential Electricity Customers by Christian Winzer and Hongliang Zhang in The Energy Journal

Footnotes

Author Contributions

Christian Winzer —Survey Methodology and Analysis,Cost-benefit analysis,Writing original draft,Review and Editing

Hongliang Zhang —Survey Methodology and Programming,Review and Editing

The authors wish to thank Karoline Gamma for her extensive support and helpful comments during the survey design. The study benefited substantially from access to the SHEDS dataset.

Funding

The work on this paper has been part of the activities of SCCER CREST,which is financially supported by Innosuisse,the Swiss Innovation Agency. The first author has also received support through SFOE grant number SI/501899-01.

All content and conclusions are exclusively the responsibility of the authors.

1.

As mentioned in section 1.2,many distribution system operators in Switzerland offer their customers either a flat tariff or a time-of-use tariff.

2.

Average of the range in (

),Abbildung 7-6 between 2.75% and 5.5%.

3.

Only applies to customers on a flat-price contract / without dual-rate meters. Installing new dual-rate meters would not make sense in view of the smart-meter roll-out and similar installation costs.

4.

Average of costs in (

),Abbildung 6-3 between CHF 145 and CHF 105.

5.

Average of costs in (

),Abbildung 6-6 between CHF 130 and CHF 230.

6.

Costs for direct load control and capacity subscriptions are higher due to the additional cost of load-control functionality.

References

Becker

Sophia

Schober

Dominik

Wassermann

Sandra

(2016). “How to Approach Consumers’ Nonmonetary Evaluation of Electricity Supply Security? The Case of Germany from a Multidisciplinary Perspective.” Utilities Policy 42(October): 74–84. https://doi.org/10.1016/j.jup.2016.06.012.

BFE (2022). “WACC–Kalkulatorischer Zinssatz.” 2022. https://www.bfe.admin.ch/bfe/de/home/versorgung/stromversorgung/stromversorgungsgesetz-stromvg/wacc-kalkulatorischer-zinssatz-gemaess-art-13-abs-3-bst-b-d.html.

Bits to Energy Lab, Ecoplan, Weisskopf Partner GmbH, and ENCO AG (2012). “Folgenabschätzung einer Einführung von «Smart Metering» im Zusammenhang mit «Smart Grids» in der Schweiz.”

Buchanan

Kathryn

Banks

Nick

Preston

Ian

Russo

Riccardo

(2016). “The British Public’s Perception of the UK Smart Metering Initiative: Threats and Opportunities.” Energy Policy 91(April): 87–97. https://doi.org/10.1016/j.enpol.2016.01.003.

Bundesrat (2021). “Bundesrat Verabschiedet Botschaft Zum Bundesgesetz Über Eine Sichere Stromversorgung Mit Erneuerbaren Energien.”June 18, 2021. https://www.admin.ch/gov/de/start/dokumentation/medienmitteilungen.msg-id-84018.html.

Burchell

Kevin

(2016). “Householder Engagement with Energy Consumption Feedback—the Role of Community Action and Communications.” Energy Policy 9. https://doi.org/10.1016/j.enpol.2015.10.019.

Buryk

Stephen

Mead

Doug

Mourato

Susana

Torriti

Jacopo

(2015). “Investigating Preferences for Dynamic Electricity Tariffs: The Effect of Environmental and System Benefit Disclosure.” Energy Policy 80(May): 190–95. https://doi.org/10.1016/j.enpol.2015.01.030 .

Chao

Hung-po

Peck

Stephen

(1997). “An Institutional Design for an Electricity Contract Market with Central Dispatch.” The Energy Journal 18(4): 85–110. https://doi.org/10.5547/ISSN0195-6574-EJ-Vol18-No1-4.

Chrzan

Keith

Orme

Bryan

(2000). “An Overview and Comparison of Design Strategies for Choice-Based Conjoint Analysis (2000).” Sawtooth Software Research Paper Series. https://sawtoothsoftware.com/resources/technical-papers/an-overview-and-comparison-of-design-strategies-for-choice-based-conjoint-analysis.

10.

Consentec, Consentec, EBP, and Polynomics (2022). “Auswirkungen einer starken Elektrifizierung und eines massiven Ausbaus der Stromproduktion aus Erneuerbaren Energien auf die Schweizer Stromverteilnetze.” https://www.newsd.admin.ch/newsd/message/attachments/74145.pdf.

11.

Consentec and Fraunhofer ISI (2018). “Optionen Zur Weiterentwicklung Der Netzentgeltsystematik.”

12.

Curtis

John

Grilli

Gianluca

Brazil

William

Harold

Jason

(2020). “Why Do Preferences for Electricity Services Differ? Domestic Appliance Curtailment Contracts in Ireland.” Energy Research & Social Science 69(November): 101705. https://doi.org/10.1016/j.erss.2020.101705 .

13.

Darby

Sarah J.

Ioana

Pisica

(2013). “Focus on Electricity Tariffs: Experience and Exploration of Different Charging Schemes.”

14.

Deller

David

Giulietti

Monica

Loomes

Graham

Price

Catherine Waddams

Moniche

Anna

Jeon

Joo Young

(2021). “Switching Energy Suppliers: It’s Not All about the Money.” The Energy Journal 42(3). https://doi.org/10.5547/01956574.42.3.ddel.

15.

Doorman

G.L.

(2005). “Capacity Subscription: Solving the Peak Demand Challenge in Electricity Markets.” IEEE Transactions on Power Systems 20(1): 239–45. https://doi.org/10.1109/TPWRS.2004.841230.

16.

Dugstad

Anders

Grimsrud

Kristine M.

Kipperberg

Gorm

Lindhjem

Henrik

Navrud

Ståle

(2021). “Scope Elasticity of Willingness to Pay in Discrete Choice Experiments.” Environmental and Resource Economics 80(1): 21–57. https://doi.org/10.1007/s10640-021-00577-7.

17.

Dütschke

Elisabeth

Paetz

Alexandra-Gwyn

(2013). “Dynamic Electricity Pricing—Which Programs Do Consumers Prefer?” Energy Policy 59(August): 226–34. https://doi.org/10.1016/j.enpol.2013.03.025.

18.

Dutta

Goutam

Mitra

Krishnendranath

(2017). “A Literature Review on Dynamic Pricing of Electricity.” Journal of the Operational Research Society 68(10): 1131–45. https://doi.org/10.1057/s41274-016-0149-4.

19.

Elcom (2021). “Elektrizitätstarife 2021.”https://www.elcom.admin.ch/dam/elcom/de/dokumente/2021/ElCom_Tarife_2021_d.xls.download.xls/ElCom_Tarife_2021_d.xls.

20.

ElCom (n.d.). “Strompreise Schweiz.” Accessed July 18, 2022. https://www.strompreis.elcom.admin.ch/?period=2021.

21.

Eniwa (2022). “Preisbestimmungen Grundversorgung 2022 Niederspannung Jahresenergieverbrauch Kleiner 50 000 KWh.” 2022. https://www.eniwa.ch/api/rm/CVV3B54T577DBJA/preisblatt-eniwa-strom-haushalt-kleingewerbe-2022-1.pdf.

22.

Fell

Michael J.

Shipworth

David

Huebner

Gesche M.

Elwell

Clifford A.

(2015). “Public Acceptability of Domestic Demand-Side Response in Great Britain: The Role of Automation and Direct Load Control.” Energy Research & Social Science 9(September): 72–84. https://doi.org/10.1016/j.erss.2015.08.023.

23.

Fell

Michael J.

(2016). Taking Charge: Perceived Control and Acceptability of Domestic Demand-Side Response. London: UniversityCollege London (UCL).

24.

Goldar

Bishwanath

Misra

Smita

(2001). “Valuation of Environmental Goods: Correcting for Bias in Contingent Valuation Studies Based on Willingness-to-Accept.” American Journal of Agricultural Economics 83(1): 150–56. https://doi.org/10.1111/0002-9092.00143.

25.

Goulden

Murray

Bedwell

Ben

Rennick-Egglestone

Stefan

Rodden

Tom

Spence

Alexa

(2014). “Smart Grids, Smart Users? The Role of the User in Demand Side Management.” Energy Research & Social Science 2(June): 21–29. https://doi.org/10.1016/j.erss.2014.04.008.

26.

Greene

William H.

Hensher

David A.

(2003). “A Latent Class Model for Discrete Choice Analysis: Contrasts with Mixed Logit.” Transportation Research Part B: Methodological 37(8): 681–98. https://doi.org/10.1016/S0191-2615(02)00046-2.

27.

Groupe-E (2021). “Tarif Unique.”2021. https://www.groupe-e.ch/sites/default/files/download-items/tarif_unique_2021_2.pdf.

28.

Hayn

Marian

(2016). Modellgestuetzte Analyse neuer Stromtarife fuer Haushalte unter Beruecksichtigung bedarfsorientierter Versorgungssicherheitsniveaus. KIT Scientific Publishing.

29.

Xiaoping

Reiner

David

(2017). “Why Consumers Switch Energy Suppliers: The Role of Individual Attitudes.” The Energy Journal 38(6). https://doi.org/10.5547/01956574.38.6.hxia.

30.

Howard

Gregory

Roe

Brian E.

Nisbet

Erik C.

Martin

Jay F.

(2017). “Hypothetical Bias Mitigation Techniques in Choice Experiments: Do Cheap Talk and Honesty Priming Effects Fade with Repeated Choices?” Journal of the Association of Environmental and Resource Economists 4(2): 543–73. https://doi.org/10.1086/691593.

31.

IWB (2022). “IWB Strom switch.” IWB. 2022. https://www.iwb.ch/Fuer-Zuhause/Strom/Stromprodukte/IWB-Strom-Switch.html.

32.

Johnson

Rich

(2000). “Understanding HB: An Intuitive Approach.” Sawtooth Software Research Paper Series15.

33.

Johnston

Robert J.

Boyle

Kevin J.

Adamowicz

Wiktor

Bennett

Jeff

Brouwer

Roy

Cameron

Trudy Ann

Hanemann

W. Michael

Hanley

Nick

Ryan

Mandy

Scarpa

Riccardo

(2017). “Contemporary Guidance for Stated Preference Studies.” Journal of the Association of Environmental and Resource Economists 4(2): 319–405. https://doi.org/10.1086/691697.

34.

Jordehi

A. Rezaee

(2019). “Optimisation of Demand Response in Electric Power Systems, a Review.” Renewable and Sustainable Energy Reviews 103(April): 308–19. https://doi.org/10.1016/j.rser.2018.12.054.

35.

Kahneman

Daniel

Knetsch

Jack L.

Thaler

Richard H.

(1991). “Anomalies: The Endowment Effect, Loss Aversion, and Status Quo Bias.” Journal of Economic Perspectives 5(1): 193–206. https://doi.org/10.1257/jep.5.1.193.

36.

Kubli

Merla

Loock

Moritz

Wüstenhagen

Rolf

(2018). “The Flexible Prosumer: Measuring the Willingness to Co-Create Distributed Flexibility.” Energy Policy 114(March): 540–48. https://doi.org/10.1016/j.enpol.2017.12.044.

37.

Lehmann

Nico

Sloot

Daniel

Ardone

Armin

Fichtner

Wolf

(2021). “The Limited Potential of Regional Electricity Marketing—Results from Two Discrete Choice Experiments in Germany.” Energy Economics 100(August): 105351. https://doi.org/10.1016/j.eneco.2021.105351.

38.

Little

Todd D.

, ed. (2013). The Oxford Handbook of Quantitative Methods. Oxford Library of Psychology. New York: Oxford University Press. https://doi.org/10.1093/oxfordhb/9780199934898.001.0001.

39.

Lopes

Marta A.R.

Antunes

Carlos Henggeler

Janda

Kathryn B.

Peixoto

Paulo

Martins

Nelson

(2016). “The Potential of Energy Behaviours in a Smart(Er) Grid: Policy Implications from a Portuguese Exploratory Study.” Energy Policy 90(March): 233–45. https://doi.org/10.1016/j.enpol.2015.12.014.

40.

Louviere

Jordan J.

Hensher

David A.

Swait

Joffre D.

Adamowicz

Wiktor

(2000). Stated Choice Methods: Analysis and Applications. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511753831.

41.

Ludwig

Patrick

Auer

Sabine

Dess

Tobias

Winzer

Christian

(2020). “NETFLEX Zwischenbericht.” https://www.aramis.admin.ch/Dokument.aspx?DocumentID=67287.

42.

Ludwig

Patrick

Winzer

Christian

(2022). “Tariff Menus to Avoid Rebound Peaks–Results from a Discrete Choice Experiment with Swiss Customers.” Energies 15(17): 6354. https://doi.org/10.3390/en15176354.

43.

Lusk

Jayson L.

Norwood

F. Bailey

(2009). “An Inferred Valuation Method.” Land Economics 85(3): 500–514. https://doi.org/10.3368/le.85.3.500.

44.

McFadden

Daniel

(1973). “Conditional Logit Analysis of Qualitative Choice Behavior.”

45.

McFadden

Daniel

(2017). “Stated Preference Methods and Their Applicability to Environmental Use and Non-Use Valuations.” In Contingent Valuation of Environmental Goods, by Daniel McFadden and Kenneth Train, 153–87. Edward Elgar Publishing. https://doi.org/10.4337/9781786434692.00012.

46.

National Grid (2018). “What Are Electricity Triads?” https://www.nationalgrideso.com/sites/eso/files/documents/Guidance%20What%20are%20Triads.pdf.

47.

Neill

Helen R.

Cummings

Ronald G.

Ganderton

Philip T.

Harrison

Glenn W.

McGuckin

Thomas

(1994). “Hypothetical Surveys and Real Economic Commitments.” Land Economics 70(2): 145–54. https://doi.org/10.2307/3146318.

48.

Nicolson

Moira L.

Fell

Michael J.

Huebner

Gesche M.

(2018). “Consumer Demand for Time of Use Electricity Tariffs: A Systematized Review of the Empirical Evidence.” Renewable and Sustainable Energy Reviews 97(December): 276–89. https://doi.org/10.1016/j.rser.2018.08.040.

49.

Nylund-Gibson

Karen

Choi

Andrew Young

(2018). “Ten Frequently Asked Questions about Latent Class Analysis.” Translational Issues in Psychological Science 4(4): 440. https://doi.org/10.1037/tps0000176.

50.

Oren

Shmuel S.

(2013). “A Historical Perspective and Business Model for Load Response Aggregation Based on Priority Service.” In, 2206–14. IEEE. https://doi.org/10.1109/HICSS.2013.39.

51.

Parrish

Bryony

Gross

Rob

Heptonstall

Phil

(2019). “On Demand: Can Demand Response Live up to Expectations in Managing Electricity Systems?” Energy Research & Social Science 51(May): 107–18. https://doi.org/10.1016/j.erss.2018.11.018.

52.

Parrish

Bryony

Heptonstall

Phil

Gross

Rob

Sovacool

Benjamin K.

(2020). “A Systematic Review of Motivations, Enablers and Barriers for Consumer Engagement with Residential Demand Response.” Energy Policy 138(March): 111221. https://doi.org/10.1016/j.enpol.2019.111221.

53.

Paterakis

Nikolaos G.

Erdinç

Ozan

Catalão

João P.S.

(2017). “An Overview of Demand Response: Key-Elements and International Experience.” Renewable and Sustainable Energy Reviews 69(March): 871–91. https://doi.org/10.1016/j.rser.2016.11.167.

54.

Potter

Jennifer M.

George

Stephen S.

Jimenez

Lupe R.

(2014). “SmartPricing Options Final Evaluation,”225.

55.

Repower (2021). “Smartpower Tarif 2021.”2021. https://www.repower.com/media/681703/tarife-2021-smartpower.pdf.

56.

Sawtooth Software Inc. (2021). “Lighthouse Studio Overview.”2021. https://sawtoothsoftware.com/lighthouse-studio.

57.

Schkade

David A.

Payne

John W.

(1994). “How People Respond to Contingent Valuation Questions: A Verbal Protocol Analysis of Willingness to Pay for an Environmental Regulation.” Journal of Environmental Economics and Management 26(1): 88–109. https://doi.org/10.1006/jeem.1994.1006.

58.

SFOE and Prognos (2020). “Energy Perspectives 2050+.” https://www.bfe.admin.ch/bfe/en/home/politik/energiestrategie-2050/dokumentation/energieperspektiven-2050.html.

59.

Sinha

Pratik

Calfee

Carolyn S.

Delucchi

Kevin L.

(2021). “Practitioner’s Guide to Latent Class Analysis: Methodological Considerations and Common Pitfalls.” Critical Care Medicine 49(1): e63–79. https://doi.org/10.1097/CCM.0000000000004710.

60.

Stenner

Karen

Frederiks

Elisha R.

Hobman

Elizabeth V.

Cook

Stephanie

(2017). “Willingness to Participate in Direct Load Control: The Role of Consumer Distrust.” Applied Energy 189(March): 76–88. https://doi.org/10.1016/j.apenergy.2016.10.099.

61.

Throndsen

William

Ryghaug

Marianne

(2015). “Material Participation and the Smart Grid: Exploring Different Modes of Articulation.” Energy Research & Social Science 9(September): 157–65. https://doi.org/10.1016/j.erss.2015.08.012.

62.

Train

Kenneth E.

(2009). Discrete Choice Methods with Simulation. Cambridge university press.

63.

Vossler

Christian A.

Doyon

Maurice

Rondeau

Daniel

(2012). “Truth in Consequentiality: Theory and Field Evidence on Discrete Choice Experiments.” American Economic Journal: Microeconomics 4(4): 145–71. https://doi.org/10.1257/mic.4.4.145.

64.

Weber

Sylvain

Burger

Paul

Farsi

Mehdi

Martinez-Cruz

Adan L.

Puntiroli

Michael

Schubert

Iljana

Volland

Benjamin

(2017). “Swiss Household Energy Demand Survey (SHEDS): Objectives, Design, and Implementation.” IRENE Working Paper. http://hdl.handle.net/10419/191509.

65.

Weller

Bridget E.

Bowen

Natasha K.

Faubert

Sarah J.

(2020). “Latent Class Analysis: A Guide to Best Practice.” Journal of Black Psychology 46(4): 287–311. https://doi.org/10.1177/0095798420930932.

66.

Whittington

Dale

Briscoe

John

Xinming

Barron

William

(1990). “Estimating the Willingness to Pay for Water Services in Developing Countries: A Case Study of the Use of Contingent Valuation Surveys in Southern Haiti.” Economic Development and Cultural Change 38(2): 293–311. https://doi.org/10.1086/451794.

67.

Wilson

Robert

(1997). “Implementation of Priority Insurance in Power Exchange Markets.” The Energy Journal 18(1): 111–23. https://doi.org/10.5547/ISSN0195-6574-EJ-Vol18-No1-5.

68.

Woo

C.K.

Kollman

Orans

Price

Horii

(2008). “Now That California Has AMI, What Can the State Do with It?” Energy Policy 36(4): 1366–74. https://doi.org/10.1016/j.enpol.2007.12.018.

69.

Woo

C.K.

Sreedharan

Hargreaves

Kahrl

Wang

Horowitz

(2014). “A Review of Electricity Product Differentiation.” Applied Energy 114(February): 262–72. https://doi.org/10.1016/j.apenergy.2013.09.070.

70.

Xiaojing

Chen

Chien-fei

Zhu

Xiaojuan

Qinran

(2018). “Promoting Acceptance of Direct Load Control Programs in the United States: Financial Incentive versus Control Option.” Energy 147(March): 1278–87. https://doi.org/10.1016/j.energy.2018.01.028.

71.

Yilmaz

Weber

Patel

M.K.

(2019). “Who Is Sensitive to DSM? Understanding the Determinants of the Shape of Electricity Load Curves and Demand Shifting: Socio-Demographic Characteristics, Appliance Use and Attitudes.” Energy Policy 133(October): 110909. https://doi.org/10.1016/j.enpol.2019.110909.

72.

Yilmaz

Selin

Chanez

Cédric

Cuony

Peter

Patel

Martin Kumar

(2022). “Analysing Utility-Based Direct Load Control Programmes for Heat Pumps and Electric Vehicles Considering Customer Segmentation.” Energy Policy 164(May): 112900. https://doi.org/10.1016/j.enpol.2022.112900.

73.

Yilmaz

Selin

Cuony

Peter

Chanez

Cédric

(2021). “Prioritize Your Heat Pump or Electric Vehicle? Analysing Design Preferences for Direct Load Control Programmes in Swiss Households.” Energy Research & Social Science 82(December): 102319. https://doi.org/10.1016/j.erss.2021.102319.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB