Environmental sensitivity questionnaire results

Get Complete Project Material File(s) Now! »

Incentives for Lowering Electricity Consumption

The principal strategies employed to incentivise households to reduce their consumption can be separated into monetary and non-monetary incentives. In this section, the dierent strategies used in the literature are described and the hypotheses that will be tested are stated.

Monetary Incentives

Monetary incentives can be separated into one of two categories: electricity cost information and pricing strategies. Monetary information is included here as although it is not a direct monetary incentive, such incentives display information in monetary terms thus informing households of how much they are spending on electricity or how much they are saving. By providing households with information as to ho much their electricity consumption costs (as opposed to information on the amount of electricity consumed) households can see the monetary benets of reducing their electricity consumption. In interviews with households participating in electricity conservation eld experiments, residents preferred to receive feedback in monetary terms as this is considered to be more relatable and more comparable than energy units (Hargreaves et al., 2010, Raw and Ross, 2011).
Further, with the installation of smart meters in residential homes, a major technological barrier to the implementation of pricing strategies such as dynamic pricing has been lifted. Dynamic pricing provides consumers with economic incentives to reduce (increase) their electricity consumption during peak (o-peak) periods by better aligning the retail price of electricity with the wholesale price in order to maintain supply and demand balance in the electricity market (Borenstein et al., 2002). Such pricing taris are eective at reducing demand during periods of high demand but are not necessarily eective at reducing overall demand (Torriti, 2012).
However, such strategies can have spillover eects when behaviour to reduce consumption during a peak period carries on into o-peak periods (Allcott, 2011a). Such pricing strategies are therefore included in the present meta-analysis.

Non-monetary Incentives

Non-monetary strategies refer to those which provide households with more detailed information on their electricity consumption. In the experimental literature, this type of incentive can be categorised into personal feedback and social feedback.

Personal Feedback

Personal feedback provides households with data on their own electricity consumption with comparisons to consumption during a dierent period, such as the previous day, month, or year. Such feedback is received in a number of ways: through detailed electricity bills (see Carroll et al., 2014, Schleich et al., 2013), online via a website or email (see Benders et al., 2006, Ueno et al., 2006, Gleerup et al., 2010, Vassileva et al., 2012, Mizobuchi and Takeuchi, 2013, Schleich et al., 2013, Harries et al., 2013, Houde et al., 2013), in real-time via a monitor in the home (see Van Dam et al., 2010, Grønhøj and Thøgersen, 2011, Alahmad et al., 2012, Carroll et al., 2014, Schultz et al., 2015).
The provision of information on individual electricity consumption allows households to develop a greater awareness of their electricity consumption. By comparing their consumption from one period to another, such information allows households to see which behaviours result in increased consumption, so that they can follow their electricity consuming activities and determine when and how they consume the most electricity, and thus when and how to reduce their consumption.

Model and Estimation Method

Meta-regression analysis is a quantitative method of systematically analysing the results of empirical studies with a common objective. It goes beyond a literature review in that it allows the analyst to calculate a mean treatment eect across studies by discovering which variables lead to dierences in experiments which study the same treatment eect (Stanley and Jarrell, 1989, Nelson and Kennedy, 2009). Metaanalyses are used to estimate a more precise estimate of the true eect of a treatment than any single study can do alone (Borenstein et al., 2009).
Using notation from Nelson and Kennedy (2009, p.8), the following meta-regression model is estimated: ~ i = 0 + 1xi1 + ::: + KxiK + ei (2.1).

Average Treatment Eects

Table 2.2 also provides both a non-weighted and weighted ATE by incentive. The ATE are weighted using study sample size as frequency weights following Schmidt and Hunter (2014) which gives more weight to studies with larger samples. The ATE across all incentives is 3.37% reduction in consumption. The weighted ATE takes into consideration the diering sample sizes in each study and equates to a 1.85% reduction in electricity consumption. This means that, on average, an incentive in a typical electricity conservation study will result in electricity savings of slightly less than 2%. In the sample of studies selected, the eect of incentives on electricity consumption ranges from an 22.2% reduction (Kendel and Lazaric, 2015) to a 13.69% increase (Torriti, 2012). From table 2.2, it can be seen that real-time feedback and monetary information have the greatest eects on electricity consumption with a weighted ATE of 2.89% and 2.86%, respectively, indicating a reduction in consumption. Pricing strategies have the smallest eect on overall electricity consumption with a weighted ATE showing a reduction in consumption of 0.99%.

Publication Bias Analysis

According to Card and Krueger (1995) there are three potential sources of publication bias in economic research: (1) a predisposition to accept studies which are consistent with the conventional view; (2) an inclination to report models based on the presence of a conventionally expected results; (3) a tendency to publish only statistically signicant results.
Potential publication bias in the sample of primary studies used in this meta-analysis can be analysed graphically using a funnel plot, as shown in g. 2.9. These graphs plot treatment eects against a measure of precision, such as the inverse standard error of the treatment eect or the square root of the sample size of the treatment group. The intuition is that the accuracy of the treatment eect increases with the level of precision. Studies with larger standard errors and smaller sample sizes are dispersed at the bottom of the graph, with the spread of treatment eects decreasing as standard errors decrease and sample sizes increase. In the absence of publication bias, the result is a symmetrical, inverted funnel shaped graph. On the other hand, if there is a publication bias, an asymmetrical funnel can result due to an absence of publications of non statistically signicant results (Egger et al., 1997, Sterne et al., 2004).

Renewable Energy and Demand Response Programmes

In the last two decades, there has been an increase in the share of renewable energy and in the number of distributed power generators (Renewable Energy Policy Network for the 21st Century, 2016). This calls for new strategies in the management of the electricity grid in order to maintain power supply reliability and quality, particularly at times when intermittent energy sources constitute a signicant part of total system capacity. This need is all the more important given that the European Union has set ambitious targets to reduce greenhouse emissions and to increase the share of renewable energy sources in the production mix by 2030 (European Commission, 2014a).
Reliable management of the electricity system requires a perfect balance between supply and demand in real time. Given the increase in renewable energy sources, this balance is harder to achieve as supply and demand levels can change rapidly and unexpectedly, in particular on high demand days and when natural conditions are unfavourable for the use of renewable energy sources. Moreover, the power generation infrastructure is highly capital intensive, such that demand side management may be one of the cheaper tools available for balancing supply and demand. Given the greater diculty of producing peak electricity, there is a need to have a more exible residential energy demand, particularly during peak periods. Demand response programmes, dened as the changes in electricity usage by end-use consumers from their normal consumption patterns in response to signals, are the main tool used or experimented in the management of the electricity grid (Balijepalli et al., 2011).
Current methods used to incentivise households to lower their energy demand include dynamic tari structures, informational incentives, or nudge-based incentives. Under certain tari structures consumers face nancial incentives to reduce their energy demand as during certain hours or on days when demand is particularly high, the price of electricity is greater than at o-peak times. This increased price is designed to induce lower electricity use at times with high wholesale market prices or when system reliability is jeopardised (Borenstein et al., 2002, Faruqui et al., 2010b,a, Hargreaves et al., 2010, Raw and Ross, 2011). Informational incentives involve providing the household with increased information on their consumption to allow them to make a more informed decision. Such incentives include information on how personal consumption compares from one day to another, or on a weekly or a monthly basis (Benders et al., 2006, Houde et al., 2013, Mizobuchi and Takeuchi, 2013, Schleich et al., 2013, Carroll et al., 2014, Schultz et al., 2015). Nudge based incentives go beyond simple information by changing the way the information is presented in order to exploit behavioural biases (Schultz et al., 2007, Thaler and Sunstein, 2008, Allcott, 2011b, Ayres et al., 2012).


Table of contents :

1 Smart meter deployment strategies across the EU-27 by 2020 as of July 2013, (European Commission, 2014b)
2 Risk-reward trade-o in dynamic pricing rates (adapted from Faruqui (2012, p.17))
3 A Home Energy Report from Opower
1 Barriers to Acceptance and Adoption of Smart Meters and Incentives to Lower Residential Energy Consumption 
1.1 Smart meter ‘Linky’ in deployment in France
1.2 Example of a TOU tari
1.3 Example of a CPP tari
1.4 Example of a PTR tari
1.5 Example of a RTP tari
2 Incentivising Households to Reduce Energy Consumption: A Metaanalysis
2.1 Geographical distribution of included studies
2.2 Treatment eects by year of publication
2.3 Treatment eects by presence of control group
2.4 Treatment eects by use of weather controls
2.5 Treatment eects by collection of socio-demographic data
2.6 Treatment eects by treatment assignment
2.7 Treatment eects by sample selection method
2.8 Treatment eects by study duration
2.9 Funnel plot of treatment eects versus sample size
3 Demand Response as a Common Pool Resource Game: Responses to Incentives to Lower Consumption 
3.1 Dynamics of average consumption by treatment
3.2 Dynamics of heating usage by treatment
3.3 Dynamics of appliance usage by treatment
4 Gain and Loss Framing of Incentives: Encouraging Individuals to Provide an Eort for Small Rewards 
4.1 A hypothetical value function (Kahneman and Tversky, 1979, p.279) .
4.2 Evolution of average number of correct tables and average diculty per period by treatment (comparison by frame)
4.3 Cumulative distribution functions of number of correct tables in each treatment
4.4 Evolution of average number of correct tables by block of 7 periods by treatment
C Appendix to Chapter 4 
C.1 Example table used in task
C.2 Stratégies de déploiement de compteurs intelligents dans l’UE-27 d’ici 2020 à compter de juillet 2013, (European Commission, 2014b)
C.3 Trade-o risque/récompense en matière de tarication dynamique (adapté de Faruqui (2012, p.17))
C.4 Une facture « Home Energy Report » d’Opower
2 Incentivising Households to Reduce Energy Consumption: A Metaanalysis
2.1 Summary of results of previous reviews and meta-analyses
2.2 Descriptive statistics and average treatment eects
2.3 Comparison of weighted average treatment eects by literature type .
2.4 Average treatment eects by study robustness
2.5 Pearson cross-correlation table
2.6 ATE correcting for publication bias
2.7 Estimation of publication bias
2.8 WLS estimation of treatment eects
3 Demand Response as a Common Pool Resource Game: Responses to Incentives to Lower Consumption 
3.1 A classication of goods
3.2 Electricity consumption choices
3.3 Number of subjects per treatment
3.4 Mean group consumption by treatment
3.5 Average group consumption (random eects estimation)
3.6 Individual consumption (random eects estimation)
3.7 Eect of feedback on individual consumption in nudge treatment
3.8 Welfare analysis at the group and the individual level
3.9 Environmental sensitivity questionnaire results
3.10 Average individual consumption by treatment and by environmental sensitivity
3.11 Altruism questionnaire results
3.12 Average individual consumption
4 Gain and Loss Framing of Incentives: Encouraging Individuals to Provide an Eort for Small Rewards 
4.1 Payos by treatment
4.2 Description of subjects per treatment
4.3 Number of correct tables overall and across all periods
4.4 Average number of correct tables by payo amount in Ex-ante and Ex-post treatments
4.5 Number of correct tables overall and across all periods
4.6 Number of correct tables across dierent stages of the game (standard deviations in brackets)
4.7 Regression estimates of average eort provision over blocks of 7 periods132
4.8 Regression estimates of eect of individual characteristics on average eort provision
A Appendix to Chapter 2 
A.1 Studies included in analysis
A.2 Reasons for studies exclusion from the analysis
A.3 OLS estimation of treatment eects
B Appendix to Chapter 3 
B.1 Number of groups by consumption level (across all periods)
B.2 Number of groups by consumption level (across all periods)
B.3 Distribution of messages received in nudge treatment by period
C Appendix to Chapter 4 
C.1 Wilcoxon rank sum tests between treatments for all periods (p-values)
C.2 Wilcoxon rank sum tests between treatments in period 1 (p-values) .
C.3 Wilcoxon rank sum tests between treatments in period 28 (p-values) .
C.4 Wilcoxon rank sum tests between treatments in period rst (p-values)
C.5 Wilcoxon rank sum tests between treatments in period last (p-values)
C.6 Wilcoxon rank sum tests of signicant dierences in eort between payo structure (p-values)
C.7 Cross-correlation table


Related Posts