can behavioural economics be used to encourage consumers ... trials...power 80% power to detect...

33
Can behavioural economics be used to encourage consumers to switch to green energy tariffs? Moira Nicolson, Gesche Huebner and David Shipworth Thursday 10 September 2015 A randomised control trial on a representative sample of British energy bill payers

Upload: others

Post on 29-Dec-2019

1 views

Category:

Documents


0 download

TRANSCRIPT

  • Can behavioural economics be used to encourage

    consumers to switch to green energy tariffs?

    Moira Nicolson, Gesche Huebner and David Shipworth

    Thursday 10 September 2015

    A randomised control trial on

    a representative sample of British energy bill payers

  • Climate change - meeting carbon emission

    targets means energy use needs to change in two ways

    1 – Reduce consumption

    2 – Adjust timing of our consumption

    (reduce consumption at the right time)

  • Time of the day

    Th

    e a

    mo

    unt o

    f e

    lectr

    icity w

    e c

    on

    su

    me

    (in

    kilo

    wa

    tts)

    0:00 12:00 24:0006:00 18:000.1

    0.7

    Source: The graph is a hand-drawn replica of a graph showing electricity load profile data (for profile class 1 – “domestic

    unrestricted customers”). Number of power stations are hypothetical.

    Energy supply needs to be reliable and

    affordable – not just clean

  • 100% renewable energy? August 2015: Jacobsen et al published roadmap for all 50 United States to

    be 100% renewable by 2050 (Energy Environ, Sci, 2015, 8)

    Photo: A wind farm in Tehachapi, California. (REUTERS/Mario Anzuoni )

  • Smart meter roll-out across world

    • Major investment – NPV of £6 billion1

    • Real time consumption measured at half

    hourly intervals

    • Key part of business case is that people

    will reduce consumption at particular

    times of day

    1 Department of Energy and Climate Change, Smart Meter Impact Assessment Final January 2014, p. 59.

  • Time period Rate you pay

    Day 7am-4pm 14p/unit

    Peak 4pm-8pm 30p/unit

    Night 8pm-7am 10p/unit

    Weekend All day 10p/unit

    Standing charge 25p per day

    Time-of-use electricity tariffs

    On this tariff, you’ll have three different electricity

    rates: day, peak and night.

  • 12Department of Energy and Climate Change, Smart Meter Impact Assessment Final January 2014, p. 59.

    To work, consumers must switch

    UK Government needs 20% uptake

    by 2030, to realise business case

    for smart meters2

    But people don’t switch• 50% haven’t switched supplier in 20

    years (Defeuilly, 2009)

    • Annual tariff switching rates are tiny,

    despite large savings (Competition & Markets Authority, 2015)

  • Traditional methods of boosting switching don’t

    seem to be working

    Regulation Making switching easier

    • Tariff Comparison Rate

    • Minimum of 4 tariffs

    • Suppliers must write to tell customers if

    they aren’t on the cheapest tariff

  • Mandates?

    Opt-outs?

  • Can we boost uptake to

    time of use tariffs without

    making them mandatory or opt-

    out?

  • “Stop harming yourself. Stop

    smoking.”

    “Not exercising regularly can

    make you gain weight”

    “Start living. Stop

    smoking.”

    “Exercising regularly can

    help you lose weight”

    “3 people die every day because

    there are not enough organ

    donors”

    “Early detection of

    HIV can prevent

    AIDS”

    “Late detection of HIV leads to

    AIDS”

    “You could save or

    transform up to 9 lives

    as an organ donor ”

    GainLoss

  • Message framing

    • Most message-framing studies are tested on health outcomes

    (Updegraff et al, 2012); message-framing on environmental

    outcomes mostly lab-based on students (following Kahneman

    and Tversky, 1981)

    • Could it influence people’s choice of energy tariff in real world

    amongst population of interest?

    • Message framing has many advantages: cheap, respects

    freedom of choice, and could be easily implemented by

    marketing staff at utility companies if we knew that a particular

    frame would be more successful

  • Tariff marketing gain-frames the financial

    benefits of switching –

    should we loss-frame them?

  • …and tell people about this?

  • The study

  • Challenge

    • Can’t measure switching rates – evidence needed now

    so electricity network operators and policymakers can

    make decisions

    • Stated preference measure using population-based

    survey experiment (Mutz, 2011)

    • Gerber and Green’s (2002) four criteria measure of

    ‘fieldiness’: authenticity of treatments, participants,

    setting, outcome

  • Design

    Power 80% power to detect effects of >10% in two-tailed test of

    significance with 95% confidence

    Participants and

    sample selection

    2020 British adult energy bill payers, members of online

    market research panel

    Recruitment Recruited by market research company to be nationally

    representative (5 demographic criteria: gender, age, region,

    social grade, working status)

    Instrument Email invitation with link to online survey (20 item

    questionnaire)

    Randomisation Random assignment to loss/gain-framed digital advert with

    and without environmental and energy security information

    Outcome measure Intention to switch to tariff on 7-point Likert scale

    Additional data

    collection

    Loss-aversion, manipulation checks, other demographic and

    household characteristics

  • Flow of study participants

  • Switch to save money

    Switch to avoid missing out on

    savings

    Switch to save money, help the environment and

    reduce risk of blackouts

    Switch to avoid missing out on

    savings, harming the environment

    and increasing risk of blackouts

    Gain Loss

    Money

    Money &

    environmental

    and energy

    security

    message

    Message-frame intervention• Participants exposed to same commercially viable time of use tariff

    • Between-subjects random assignment to message-frame (1:1 allocation procedure)

    • Randomisation carried out on rolling basis by survey software use Excel’s random

    number generator

  • Scenario used to measure consumer interest in demand-side response

    in Spence et al (2015), Nature Climate Change, 5, 550-554.

  • Measured loss-aversion-50/50

    gamble questionsGamble Yes I would

    take this

    gamble

    No, I would

    not take this

    gamble

    #1 If the coin turns up heads then you lose £2; if the coin turns up

    tails then you win £6 o o

    #2 If the coin turns up heads then you lose £3; if the coin turns up

    tails then you win £6 o o

    #3 If the coin turns up heads then you lose £4; if the coin turns up

    tails then you win £6 o o

    #4 If the coin turns up heads then you lose £5; if the coin turns up

    tails then you win £6 o o

    #5 If the coin turns up heads then you lose £6; if the coin turns up

    tails then you win £6 o o

    #6 If the coin turns up heads then you lose £7; if the coin turns up

    tails then you win £6 o o

    Source: Loss-aversion questions the same as used in Gachter et al (2010),

    adapted from Kahneman and Tversky (1979) to make suitable for online survey.

  • Analysis• Pre-analysis plan

    • Treatment effects estimated using OLS linear regression with

    and without select baseline covariates (Ordered Logit

    robustness check)

    • Heterogenous treatment effects

    - Pre-payment customers

    - Loss-averse (continuous, categorical)

    - Infrequent switchers

    • Highly specified model to investigate what variables are

    statistically significantly correlated with intention to switch to a

    time of use tariff

    • Multiple comparisons correction applied throughout (Benjamini

    and Hochberg [1995]) method)

  • Results

  • Randomisation and manipulation checks

    • Randomisation check on baseline characteristics revealed covariate balance across groups was excellent (no more differences than expected by chance)

    • Manipulation checks: participants perceived framing manipulation as intended and no differences in evaluation of messages across groups (e.g. ease of understanding, knowledge test)

    • Attrition – 14 dropped out after randomisation but prior to responding to outcome measure (0.7% attrition rate)

    • Average time taken to complete survey – 30 minutes (20 items)

  • 1: High variation in willingness to switch to a

    time of use tariff (outcome measure)

    0

    .05

    .1.1

    5.2

    .25

    1 2 3 4 5 6 7

    Pro

    port

    ion o

    f consum

    ers

    Stated willingness to switch to tariff on Likert scale

  • 2: British energy bill payers are loss-averse

    Percent accepted

    Implied

    acceptable

    loss

    Unweighted

    proportion

    (standard

    deviation)

    Weighted

    proportion

    (standard

    error)

    Accept all lotteries 0.05 (0.22) 0.05 (0.01) £7

    Accept lotteries #1-#5, reject lottery #6 0.02 (0.16) 0.02 (0004) £6

    Accept lotteries #1-#4, reject lotteries #5-#6 0.03 (0.17) 0.03 (0.005) £5

    Accept lotteries #1-#3, reject lotteries #4-#6 0.07 (0.25) 0.07 (0.01) £4

    Accept lotteries #1-#2, reject lotteries #3-#6 0.23 (0.42) 0.24 (0.01) £3

    Accept lottery #1, reject lotteries #2-#6 0.24 (0.43) 0.23 (0.01) £2

    Reject all lotteries 0.28 (0.45) 0.28 (0.01) £

  • 12

    34

    56

    7

    Mean s

    tate

    d w

    illin

    gn

    ess t

    o s

    witch

    Increasing consumer loss-aversion

    3: Loss aversion was statistically significantly negatively

    correlated with willingness to switch at 99.9% level

    95% of British

    population

    30% of British

    population

    Notes: Statistical significance testing performed using OLS linear regression, with survey weights and correction for multiple comparisons (Benjamini and

    Hochberg [1995], with and without region fixed effects. Ordered Logit robustness checks. Covariates included: a continuous measure of loss-aversion (0-6

    ascending); gender; education; tenure; age; employment status; household occupancy patterns; interruptible electric goods ownership; children at home;

    switched tariff this year, last year, ever; existing time of use tariff customer. Alpha = p

  • 4 – Message framing results

  • 4: No statistically significant difference in willingness to switch

    across experimental groups

    Experimental group

    Me

    an

    will

    ing

    ne

    ss t

    o s

    witch

    to

    ta

    riff o

    n L

    ike

    rt s

    ca

    le

    Error bars represent standard error around mean intention to switch by experimental group. p-values ranged

    between 0.448-0.840 across OLS and Ordered Logit robustness checks.

    2.88 2.71 2.89 2.900.00

    1.00

    2.00

    3.00

    4.00

    5.00

    6.00

    Money (gain) Money (loss) Money &environment/security

    (gain)

    Money &environment/security

    (loss)

    7

    1

    2

    3

    4

    5

    6

  • 5 – Causal chain of process: no

    sub-group effects

    • Loss-framing had no effect on loss-averse

    (continuous/categorical)

    • Environmental framing had no effect on

    hybrid vehicle owners (not in PAP)

    Notes: Statistical significance testing performed using OLS linear regression, with survey weights and correction for multiple

    comparisons (Benjamini and Hochberg [1995], with and without region fixed effects. Ordered Logit robustness checks. The treatment

    dummy was created by pooling participants assigned to groups 1 and 3 (loss-framed messages) into one large treatment group and

    pooling participants assigned to leaflets 0 and 2 (gain-framed messages) into one large control group. Alpha = p

  • Conclusions• British energy bill payers are loss-averse and this is negatively correlated with

    willingness to switch to time of use tariff tested – priming effects? Measurement error?

    • No influence of message framing (80% power, balance checks fine)

    • Framing was too weak? But manipulation checks fine and if too different it isn’t framing

    • Loss-framing, with or without environmental message, either:

    o Won’t work to boost uptake rates to time of use tariffs – no treatment effect OR;

    o It won’t work very well – small treatment effect in order of

  • Thank you –

    any questions/feedback?

    Moira Nicolson

    [email protected]

    With thanks to Engineering and Physical Sciences Research Council (EPSRC) for

    doctoral training grant and to Smart Energy GB, for supporting the research.