Daylight Time and Energy Evidence from an Australian Experiment

1
Daylight Time and Energy Evidence from an
Australian Experiment
Ryan Kellogg
Hendrik Wolff
Presentation October 2007 Department of
Economics - University of Washington
2
Daylight Saving Time (DST)A tool for energy
conservation?

• Several countries debate about DST
• - USA Energy Policy Act of 2005
• Title I Extends DST by four weeks to reduce
  energy consumption by the equivalent of 100,000
  barrels of oil for each day of the extension.
• - Australia, UK, Japan, Korea and others are
  considering extending DST to curb GHG emissions

3
We use a quasi-experiment to assess whether
extending DST saves electricity

• 2000 Sydney Olympic Games
• - two states in Australia began to observe DST
  two months earlier than usual
• Preview of main result
• Contrary to previous DST-literature Extending
  DST does not save electricity

4
(No Transcript)
5
I say it is impossible that people lived so long
by the smoky, unwholesome, and enormously
expensive light of candles, if we had as much
pure light of the sun for nothing.
Benjamin Franklin, Paris, 1784
6
I say it is impossible that people lived so long
by the smoky, unwholesome, and enormously
expensive light of candles, if we had as much
pure light of the sun for nothing.
Benjamin Franklin, Paris, 1784 Misallocation
causes Paris to consume an additional 64
million pounds of tallow and wax annually.
7
(No Transcript)
8
(No Transcript)
9
Artificial Lighting
Sleep
10
(No Transcript)
11
(No Transcript)
12
Prior studies find extending DST into March
saves electricity by 0.6-3.5

Studies are directly used by Governments to
decide on DST extensions USA 1 New
Zealand 3.5 Australia 1-3.5
California 0.6 Ontario, Canada
2.2 However, studies are based on simulations
and extrapolations, rather than empirical
evidence
13
Nixon 1973 Emergency DST Energy Conservation Act

• Observational Study finds 1 savings based on
  73-75 DST extension in the USA
• However
• technology changed
• potentially confounded

14
Yellow Regions observing DST as of 2006
15
Road Map

• Background on the event and graphical results
• Treatment effect estimation
• Testing prior DST simulation models
• Conclusions

16
Typically 3 states observe DST

• Typically SA, NSW, and VIC observe DST from
  October to March

Northern Territory
Queensland
Western Australia
South Australia (SA)
New South Wales (NSW)
VIC
Victoria
17
In 2000 NSW and VIC are treated with 2 months
extended DST

• Typically SA, NSW, and VIC observe DST from
  October to March
• In 2000
• NSW and VIC start DST two months earlier

Northern Territory
Queensland
Western Australia
South Australia (SA)
New South Wales (NSW)
VIC
Victoria
18
Policy change was not prompted by intent to
conserve energy

• Cited rationales
• Fewer shadows on the fields improve TV
  broadcasting
• Shift visitors between stadia in daylight

Northern Territory
Queensland
Western Australia
South Australia (SA)
New South Wales (NSW)
VIC
Victoria
19
Policy change was not prompted by intent to
conserve energy

• Cited rationales
• Fewer shadows on the fields improve TV
  broadcasting
• Shift visitors between stadia in daylight

Northern Territory
Queensland
Western Australia
South Australia (SA)
New South Wales (NSW)
Control State
VIC

• Olympic events confound NSW
• International tourism
• Construction activities
• ? Treated state VIC, Control state SA

Adelaide
Melbourne
Treated State
20
Time Line in 2000
August 27 DST starts in VIC
October 29 DST starts in SA
September 15 October 1 Sydney Olympics in NSW
21
Time Line in 2000
August 27 DST starts in VIC
October 29 DST starts in SA
VIC Treatment I
VIC Treatment II
September 15 October 1 Sydney Olympics in NSW

• The Olympic events present a potential confound
  in VIC
• Increased TV ratings, Carnival events around
  public mega screens
• Drop Olympic period from treatment period in VIC

22
SA (control state)shows no effect of DST
extension

Average half hourly electricity demand during the
treatment period
23
Average half hourly electricity demand during the
treatment period
VIC (treated state) shows regular load pattern
in control years
24
VIC 2000shows intra-day shift in electricity
load
25
Dataset 1999-2005 panel of

• half hourly electricity demand wholesale prices
• hourly weather
• Temperature, Precipitation, Wind, Pressure,
  Sunshine, Humidity
• Day of week, school-vacation, holidays,
  transition vacation days
• Employment, Gross-State-Product, population

26
Difference in Differences Mechanics

• Control structure is two-fold
• (a) spatial across states
• controls for differences between states
• (b) temporal over years
• controls for any shock on the national level

27
Standard DID
28
Augment standard DID model by estimating
triple-DID Treatment Effect Model

• Triple-DID control structure is three-fold
• (a) Spatial across states
• (b) temporal over years
• (c) temporal within days using early afternoon
  hours k 1200-1430 as within controls
  
• With triple-DID we dont depend on
• nearby months and
• Seasonal variations
• Model robust against shocks affecting the level
  of any day
• (demand is
  function of 1800 variables)

29
Half hourly treatment effects of extending DST on
electricity use
The estimated effect of extending DST in VIC,
disaggregated by half-hour, with 95 confidence
intervals. Standard errors are clustered by day.
30
Testing the Electricity Saving HypothesisPercent
age change due to DST
New Zealand
31
Testing the Electricity Saving HypothesisPercent
age change due to DST
Canada (Ontario)
New Zealand
32
Testing the Electricity Saving HypothesisPercent
age change due to DST
USA
Canada
New Zealand
33
Testing the Electricity Saving HypothesisPercent
age change due to DST
Canada
New Zealand
California
USA
34
Testing the Electricity Saving HypothesisPercent
age change due to DST
Canada
New Zealand
Australia
USA
California
35
Testing the Electricity Saving HypothesisPercent
age change due to DST
Australia
Canada
New Zealand
USA
California
36
Testing the Electricity Saving HypothesisPercent
age change due to DST
weekends
weekdays
Canada
New Zealand
USA
California
37
Value of results

• First quantitative study to show that DST does
  not decrease electricity consumption,
  contradicting prior research
• Australia considers extending DST to cut GHG
  emissions
• Can we transfer the results to the U.S.?
• San Francisco latitude climate similar to
  Melbourne
• Next we examine the California Simulation model

38
California identifies three benefits of extending
DST

• Extended DST
• (1) saves non-renewable resources
• Electricity use decrease
• (2) increases consumer welfare
• Reduction in peak evening demand price decreases
  
• CA benefits up to 1.3 billion annually
• (3) helps to avoid extreme events
• Likelihood of blackout decreases

39
CEC 2001 Simulation
40
Simulation unable to predict intraday-change in
demand
41
Treatment in Victoria 2000 leads to a price spike
MW
42
Conclusions I

• ...none of the three benefits of extended DST
  could be confirmed. The extension
• - does not save electricity
• - increases prices
• - increases the critical demand spikes
• findings are of policy interest since Australia
  considers re-introducing the 2000-DST-schedule to
  curb GHG emissions
• results suggest U.S. will not benefit from
  extending DST
• extending DST is not a quick fix for energy
  conservation.

43
Conclusions II
U.S. Energy Bill of 2005 If the study does not
report adequate savings, Congress should
consider to return to the original Daylight
Savings Time schedule. Policy Recommendation
.Pull out DST legislation from Energy
Bills .Health Benefits? Business?
44
Penetration trends of air conditioning by state
45
Characteristics of Generators
Two Issues - Is Real Time Pricing Green?
Impacts of Demand Variance (Holland Mansur,
2004) - Forecasting Error, DST learning
46
National Electricity Market, Australia
47
Settlement of electricity prices in VIC, NSW, QLD
and SA
48
?? demand ?? price
M Tu W Th F Sa Su M T W T F
M T W T F
------------- School Vacation ----------
49
Table 2 Summary statistics of data used from
1999 to 2001, 27 August to 27 October
50
(No Transcript)
51
Selection of the afternoon hours
52
(No Transcript)
53
(No Transcript)
54
(No Transcript)
55
Equator
56
Melbourne
57
Melbourne
58
Artifical lighting
Sleep
59
(No Transcript)
60
(No Transcript)
61
Sunshine Hours during Standard Time September,
Melbourne
62
Sunshine Hours during DST extension September,
Melbourne
63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
(No Transcript)
68
(No Transcript)
69
San Francisco
70
Melbourne