Energy Security in Singapore

Analyzing Elasticity Trends for
Singapore Household Electricity
Demand – Implications for Policy
Making and the Rebound Effect
Allan Loi, Ng Jia Le
34th United States Association for Energy Economics Conference
Hyatt Regency Tulsa
Tulsa, Oklahoma, USA
25 October 2016
2
Overview
• Introduction - Singapore’s Household
Electricity Needs
• Motivation and Objectives of Study
• Methodology and Specification
• Empirical Results
• Policy Implications
3
Introduction - Singapore’s Household
Electricity Needs
• Small City-State. Approximately 760 km2
• Population density: 8000 people/m2
• 1 degree North of equator, temperatures at 25oC during
nighttime, 32oC during daytime
• 100% urbanized and electrified since 1990
• Fixed electricity prices for residential consumers, set by
Singapore Power Services Ltd.
• Retail Contestability available for residential consumers in
2018
• Mandatory Energy Labelling Scheme (2008). Mandatory
Energy Performance Standards (2011)
4
Introduction - Singapore’s Household
Electricity Needs
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
Central Region
319.4
317.8
319.7
312.7
327.2
333.0
323.2
324.7
328.4
331.2
East Region
410.6
404.9
405.0
393.0
404.9
415.6
401.7
401.9
404.3
403.7
North East Region
375.0
373.4
374.6
366.6
384.9
391.8
372.7
371.7
372.4
372.2
North Region
386.5
384.9
383.9
379.1
394.5
403.8
392.1
392.7
397.7
388.2
West Region
379.0
375.2
374.0
366.6
383.1
389.7
376.5
375.8
377.9
379.1
5
Introduction - Singapore’s Household
Electricity Needs
• Electricity is a necessity for 1-2 room flats
• Condo dwellers face decreasing electricity demand over time. More efficient?
• In 2011, Singapore’s temperature dipped by a lot, resulting in large decreases
in electricity.
6
Motivation and Objectives of Study
• To construct an electricity demand model that accounts
for more heterogeneity.
(regions and dwelling types)
• Better understand the role of price asymmetry in
electricity demand for Singapore, which can have
implications for rebound effect and demand response.
• Provides a starting point to further investigate the
rebound phenomenon via disaggregated, individual
household data.
7
Motivation and Objectives of Study
• There are many cities in the US that may exhibit similar
demand characteristics typical of a tropical region.
Hawaii, Denver, Tulsa, Miami Beach, Santa Barbara etc…
• Also, non studies that analyze within-city electricity
demand that take into account heterogeneity across
different classifications of household electricity use on a
National level are rare (excludes randomized control
trials).
8
Methodology and Specification
Panel data over 10 years (2005 to 2014) on average
household electricity demand, income, prices, and
dummies pertaining to recession (2008) and EE policy.
All data are either from the National Environment Agency,
or from the Singapore Department of Statistics.
9
Methodology and Specification
We use a Multivariate Panel Specification, utilizing both
one-way cross-sectional fixed effects, and the fully
modified least squares estimator (FMOLS).
Across Dwelling Types:
• ln Ei,t = α0i + α1 ln Pt + α2 ln HSi,t + α3 ln Ii,t + α4Rt
+ α5 T + α6 EEt + α7 ln Ei,(t-1) + α8Temp + εi,t
Across Districts:
• ln Ei,t = boi + b1 ln Pt + b2 Rt + b3 T + b4 EEt
+ b5 ln Ei,(t-1)+ εi,t
Across Districts, income, temperature and household size
are not used as they are not available across regions for the
panel period (only for 2013).
10
Methodology and Specification
Price Asymmetry:
Decomposition of yearly tariffs based on Gately et al (2002) and
Wolfram (1971). Price decomposition methods were first developed to
study agricultural commodities, but found use in the energy sector in
recent times.
11
Methodology and Specification
Price Asymmetry:
Decomposition of yearly tariffs based on Gately et al (2002) and
Wolfram (1971).
12
Empirical Results
Selected Price and Income Estimates across dwellings:
(After testing for each series as an I(1) process based on)
C
Fixed Effects
(1)
-0.309
Log(Et-1)
0.830***
log(Pt)
-0.088**
log(household_Size) 0.230***
(2)
FMOLS*
(3)
(3A)
0.732***
0.786***
0.736***
-0.062***
0.196***
-0.186***
0.207***
‘-0.159***
0.236***
0.148***
0.142***
-0.031***
‘-0.03***
0.418**
log(income)
0.154***
0.184***
recession
-0.029***
-0.035***
DUMee
-0.051***
-0.059***
Log(temp)
* recession dummy entered as an exogenous regressor
in FMOLS specification
13
Empirical Results
Selected Price and Income Estimates across Regions
(After testing for each series as an I(1) process)
C
Log(Et-1)
log(Pt)
recession
One Way Fixed Effects
(4)
(5)
2.651***
1.962***
0.615***
0.739***
(FMOLS)
(6)
0.418***
-0.126***
-0.138***
-0.122***
trend
0.008***
DUMee
-0.031***
1 Specification uses cross section
GLS weights
0.008***
-0.031***
Yes
-0.025***
14
Empirical Results
• Results show electricity demand inelastic to price
(-0.05 to -0.235)
• Inelastic response to income, although it plays a bigger
role affecting demand (0.17 to 0.21)
• Household size most influential for demand usage
• Energy Efficiency could have led to 6% reduction in
electricity usage as in Model (6)
These results are consistent with an earlier work done in
Singapore by Loi and Loo (2016) for time series between
1980-2014.
15
Empirical Results
Table 5: Price Asymmetry for household electricity use - Panel Estimates
Gately Decomposition
(3A)
(3B)
(3C)
Wolfram Decomposition
(3D)
Log(Et-1)
0.790*** 0.523*** 0.903*** 0.635***
log(Pmax)
-0.158*** -0.373*** -0.003
log(Pinc)
-0.220*** -0.050*
log(Pdec)
-0.476*** -0.460*** -0.482*** -0.454***
0.024**
0.022*
-0.271*** -0.052
0.008
-0.028
recession
DUMee
Wald Test
-0.031*** Yes
1%
-0.026*** Yes
1%
(3F)
(3G)
(3H)
(3I)
0.548***
0.548*** 0.734*** 0.769*** 0.776***
-0.068***
0.005
-0.180***
-0.291*** -0.194*** -0.346*** -0.477***
0.294***
0.272*** 0.197*** 0.188*** 0.185***
-0.111
log(household) 0.186*** 0.395***
log(income)
(3E)
1%
-0.125*** -0.051*** -0.214***
0.120***
0.150***
0.0378***
Yes
Yes
-0.03*** -0.016*** No
Yes
Yes
-0.056*** -0.038*** -0.033***
1% %
1%
1%
1%
1%
Note: All the Specifications are based on the FMOLS Specification, with heterogeneous
first-stage coefficients. In addition, no trends were specified across all the
models
• People respond more towards price decreases
• Surprising result if you consider individuals to be risk-averse towards costs.
1%
16
Empirical Results
• Price responsiveness higher for price decreases than
increases.
• Insertion temperature data does not change the
comparative aspects of price asymmetry.
• Estimates range between -0.18 to -0.482.
17
Policy Implications
• Likely low responsiveness towards demand response
when market opens for retail contestability. – Low price
& income elasticities.
• More outreach is needed to educate consumers on the
benefits of demand response and switching.
• Comfort has a greater importance on consumers’ welfare
than cost as income level increases. – May not be
worthwhile sacrificing thermal pleasures for cost
savings.
• Price asymmetry exists, and hence the rebound may
follow similar trend. --- Magnitude needs to be further
estimated with individual household data at higher
frequencies.
18
Thank you!
Energy Studies Institute
29 Heng Mui Keng Terrace
Block A, #10-01
Singapore 119620
Loi Tian Sheng, Allan
Tel: (65) 6516 2349
Email: [email protected]
19
Singapore’s Temperature