The Incidence of a Carbon Tax across Thai Household Groups: the

Residential Energy Demand and Behaviour:
its Implication for Climate Mitigation
Policies and Energy Subsidy Reform in
Thailand
By Supawan Saelim
1
School of Development Economics,
National Institute of Development Administration, Thailand
Presented at 40th IAEE International Conference, Singapore
19 June 2017
Agenda
2
Introduction
Methodology
Data
Results
Conclusion and policy implications
3
Introduction



Motivation
Research Questions and Objectives
Household consumption
Motivation - the need for policies
4
Current
Challenges
Carbon tax/
Energy subsidy
removal
Submitted INDCs to reduce GHG emissions
by 20 to 25% below BAU by 2030
Existing distortion in fuel price structure
and the downward trend of energy prices
Substantial budget needed to finance
renewable and alternative energy policies
Motivation - Distributional concerns
5
Policies

&
Its
distribution
al concerns

A market-based approach and costeffective migration policy to achieve
environmental goals and raise
revenues i.e., carbon tax
But, there are negative
distributional concerns on
 Regressive welfare effects
 Worsen income inequality and
poverty reduction
 Raise public concerns and political
resistance
Motivation - Distributional studies
6
Progressive results
Regressive results

UK: Food and energy are necessities for
poor households (Symons et al., 1994)

Sweden: Higher energy demand for rural
households (Brannlund and Nordstrom,
2004)


Denmark: Low-income households
proportionally consume more food and
public transport while rural households have
high demand for heating, electricity and
transport (Weir et al., 2005)
Ireland: Carbon tax is more regressive for
home heating than for motor fuels,
reflecting the necessity of home heating for
the poor (Callan et al., 2009)

Italy: Carbon tax mainly hit transport
fuels, so poor households are less
affected due to their low car ownership
(Tiezzi, 2005)

Indonesia: Consumption baskets for poor
households are less-energy incentive,
typically those living in rural areas (Yusuf
and Resosudarmor, 2015)

China: Urban households spend more on
energy and high carbon-intensive goods
while rural households spend a larger
share on food (Brenner et al., 2007)
Household energy consumption plays a major role in influencing the
distributional results of carbon tax impact.
Research questions and objectives
7
Q1: Is pricing policy effective in reducing energy consumption in the
residential sector?
Q2: Which household groups are more responsive to price changes?
Effectiveness
Equity
Social
Policy
design
Objective: To empirically estimate energy demand in the residential
sector and calculate price and expenditure elasticities
Household energy consumption
8

Energy expenditure accounts for 11% of average monthly
household expenditure. About 94% of monthly energy expenditure
mainly consists of transport fuels (67%) and electricity (27%)
2013 Share of average monthly expenditure
Vehicle
10%
Housing/
Appliance
19%
Others
26%
Energy
11%
Food/Drink/Tobacco
34%
Source: Thai Household Socio-Economic Survey (SES)
Transport
fuels
7%
Electricity
3%
other energy
1%
Household energy consumption
9
Panel A: By income class
11.0%
9.0%
7.0%
7.7%
8.3%
8.4%
8.8%
6.1%
5.0%
3.8%
3.6%
3.6%
3.4%
3.2%
3.0%
1.0%
Oil
Quintile 1
Quintile 2
Panel B: By region
11.0%
9.6%
9.0%
7.0%
Electricity
Quintile 4
Quintile 5
Quintile 3
7.9%
8.1%
7.6%
5.9%
4.3%
5.0%
4.0%
3.6%
3.0%
1.0%
Oil
Bangkok
Electricity
Central
North
Source: Thai Household Socio-Economic Survey, 2013
Northeast
South
3.0%
3.4%
10
Methodology
Demand estimation model
Energy Elasticities
I. Demand Estimation
11

The Quadratic Almost Ideal Demand System (QUAIDS)
model (Bank et al., 1999)
𝑤𝑖 = 𝛼𝑖 +
𝑘
𝑗=1 𝛾𝑖𝑗
where ln𝑎(𝑝) = 𝛼0 +
𝑙𝑛 𝑝𝑗 + 𝛽𝑖 𝑙𝑛
𝑘
𝑖=1 αi ln 𝑝𝑖
+
𝑚
𝑎(𝑝)
1
2
𝑘
𝑖=1
+
𝜆𝑖
𝑏(𝑝)
𝑙𝑛
𝑘
𝑗=1 γij 𝑙𝑛𝑝𝑖
𝑙𝑛𝑝𝑗 ;
𝑘
𝛽
𝑝𝑖 𝑖
b(p) =
and
λ (𝑝) =
𝑘
𝑖=1 λi ln 𝑝𝑖
𝑖=1
Theoretical restrictions: additivity, homogeneity and symmetry
k
i=1 αi
= 1,
k
i=1 βi
= 0,
k
j=1 γij
=0
k
i=1 λi
𝑚
𝑎(𝑝)
= 0, and γij = γji
2
I. Demand Estimation
12
 Assume two-stage budgeting process
Consumption
Non-durable
Electricity
(w1)
Transport fuels
(w2)
Food and
beverage
(w3)
Durable
Other nondurable
(w4)
I. Demand – Econometric Model
13

Econometric model specification:
• Incorporate demographic
variables through intercept
term (Translating approach)
• Correct for endogeneity
using IV and augmented
regression techniques
• Iterated Linear Least Square estimator
(ILLE) estimation techniques imposing
theoretical restrictions
𝑘
wi = 𝛼𝑖 (𝑧) +
𝑗=1
𝑚
𝜆𝑖
𝑚
γij ln 𝑝𝑗 + 𝛽𝑖 ln
+
ln
𝑎(𝑝, 𝜃)
𝑏(𝑝, 𝜃)
𝑎(𝑝, 𝜃)
2
+ 𝜌𝑖 𝑣 + 𝜖𝑖
I. Demand – Elasticity measures
14

Use parameters obtained from demand system estimation to
calculate elasticities of demand
Expenditure
elasticity
Uncompensated
price elasticity
Compensated
price elasticity
𝛆𝐦 = 𝛍𝐦 /𝐰𝐢 + 𝟏
where μm =
𝜕 wi
𝜕 ln m
= βi +
2λi
b(p)
ln
m
a(p)
𝛆𝐮𝐢𝐣 = 𝛍𝐢𝐣 /𝐰𝐢 - 𝛅𝐢𝐣
where μij =
𝜕 wi
𝜕 ln pj
= γij − μm ( αi +
k γjk ln pk)
λi βj
−b
p
ln
m
a(p)
𝛆𝐜𝐢𝐣 = 𝛆𝐮𝐢𝐣 + 𝒘𝒋 𝛆𝐦
which is derived from the Slutsky equation.
Income and price elasticity matter in explaining the extent of change in
consumption as a result of change in price
2
15
Data
I. Household survey data (SES)
II. Consumer price indices (CPI)
I. Household survey data
16

The national Thai Household Socio-economic Survey (SES) for
consumption and income survey conducted by The National
Statistical Office (NSO)

Data for demand estimation
 SES for the year 2009, 2011 and 2013 (128,665 observations)
 Pooled cross-sectional dataset was reduced to 114,470
observations to ensure the correspondence between the
expenditure and price data
Distribution of observations
17
Dimension
Month of interview
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Total
Survey year
Y2009
3,656
3,694
3,653
3,604
3,642
3,668
3,679
3,670
3,563
3,671
3,660
3,684
43,844
Total
Y2011
3,511
3,481
3,517
3,518
3,521
3,470
3,469
3,525
3,588
3,469
3,483
3,531
42,083
Y2013
3,524
3,612
3,525
3,513
3,555
3,525
3,593
3,528
3,540
3,562
3,590
3,671
42,738
10,691
10,787
10,695
10,635
10,718
10,663
10,741
10,723
10,691
10,702
10,733
10,886
128,665
II. Price indices
18

Monthly regional consumer price indices are conducted
by Bureau of Trade and Economic Indices, Ministry of
Commerce. The indices are calculated based on the
modified Laspeyres’ formula
𝑃𝐼 =
𝑝𝑖𝑡
𝑛
𝑖 𝑤𝑖0 𝑝
𝑖0
𝑛
𝑖 𝑤𝑖0
∗ 100
where 𝑤𝑖0 is a weight in the reference period (i.e., year
2011).
 This formula is also applied to calculate price index for
aggregation of other non-durable consumption.
II. Description of variables
19
Variables
lny
lnx
w1
w2
w3
w4
lnp1
lnp2
lnp3
lnp4
nyg
noyg
ownveh
oele
male
noneco
farm
employ
bus
BKK
C
N
NE
S
Type
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Dummy
Reference
Dummy
Dummy
Dummy
Reference
Dummy
Dummy
Dummy
Dummy
Description
Log of monthly disposable income
Log of monthly non-durable expenditure
Share of electricity
Share of transport fuels (Tfuels)
Share of food and non-alcohol beverages (FB)
Share of other non-durable expenditure (Others)
Log of price indices for electricity
Log of price indices for transport fuels
Log of price indices for food and non-alcohol beverages
Log of price indices for other non-durable goods and services
Numbers of young members with age less than 15 years old
Numbers of adult members with age more than 15 years old
Numbers of owned vehicles i.e., automobile, pickup and van
Numbers of owned large electrical appliance
Male headed (=1); Female headed (=0)
Households mainly earn income from non-economic activities (=0)
Households mainly earn income from farm operation (=1)
Households mainly earn income from wages and salaries (=1)
Households mainly earn income from non-farm business (=1)
Living in the Bangkok and metropolitan area (=0)
Living in the Central region (=1)
Living in the North region (=1)
Living in the Northeast region (=1)
Living in the South region (=1)
20
Results
I. Energy Elasticities
II. Elasticities across income groups
Residential demand elasticities
21


Energy demand is inelastic to changes in energy prices.
Transport fuels are more price-elastic than electricity
Energy
types
Predicted
budget share
Electricity
3.6%
Tfuels
7.8%
Elasticities
Expenditure
Price (𝜺𝒖 )
0.634***
(0.0070)
1.001***
(0.0090)
-0.526***
(0.0130)
-0.602***
(0.0210)
Standard errors in parentheses
Elasticities across income groups
22

Energy demand is more necessity for higher-income households

High-income households are more responsive to energy price changes
Energy types
Income
groups
Low
Electricity
Mid
High
Low
Tfuels
Mid
High
Elasticities
Expenditure
Price (𝜺𝒖 )
0.656
0.623
0.603
1.311
0.936
0.735
-0.521
-0.522
-0.531
-0.514
-0.596
-0.647
Income groups: low (Quintile 1 and 2), Mid (Quintile 3 and 4), High (Quintile 5)
23
Conclusion and policy implications
Residential energy demand
Implication for distribution analysis
Residential energy demand
24





Residential energy demand is price inelastic.
Households are more responsive in reducing the consumption of
transport fuels as compared to electricity when price changes.
Households at higher income distribution are more responsive to
changes in energy prices
More flexible functional form (i.e., quadratic) for residential energy
demand is needed to capture the observed consumption patterns for
future demand forecasting or projection of emission levels.
Other measures than pricing policy such as investment in renewable
energy and improvement of public transports are needed to allow
households to have more choices to alter their consumption.
Potential distributional results
25

Potential biases for the distributional results of energy and
climate policies when assuming uniform elasticity in the
residential energy sector

An increase in energy prices induced by carbon tax or energy
subsidy reform is likely to be progressive in Thailand as highincome households consume relatively larger energy
consumption.

High-income households are more vulnerable to reduce
energy consumption in response to an increase in energy
prices induced by the policies than low-income households.

The characteristics of consumers and their behavior matter in
determining the distributional incidence of environmental and
energy policies.
26
Thank you
Appendix: Significance of variables
27
Variable
Null hypothesis
lambda_lnx2
No quadratic term
rho_vexp
Expenditure exogeneity
alpha_nyg
chi2 (3)
Prob>chi2
3787.12
0.000
396.22
0.000
No effects of number of young members
6669.12
0.000
alpha_notyg
No effects of number of adult members
6362.46
0.000
alpha_ownveh
No effects of number of owned vehicles
29741.91
0.000
alpha_oele
No effects of number of owned electrical appliance
12862.92
0.000
alpha_male
No effects of male-headed household
843.27
0.000
alpha_farm
No effects of farmers
3504.28
0.000
alpha_employ
No effects of employees
3370.56
0.000
alpha_bus
No effects of self-employed
1305.63
0.000
alpha_C
No effects of living in the Central region
1316.44
0.000
alpha_N
No effects of living in the North region
2958.05
0.000
alpha_NE
No effects of living in the Northeast region
4411.57
0.000
alpha_S
No effects of living in the South region
3147.41
0.000
Appendix: Robustness
28
Table B4 Robustness of elasticities to alternative data set
Compensated price
elasticities
Panel A: All households
Electricity
Tfuels
FB
Others
Report
-0.503***
(0.0130)
-0.524***
(0.0210)
-0.508***
(0.0040)
-0.533***
(0.0060)
Alternative dataset
A
B
C
-0.494***
(0.0140)
-0.522***
(0.0200)
-0.511***
(0.0040)
-0.529***
(0.0060)
-0.490***
(0.0130)
-0.527***
(0.0200)
-0.520***
(0.0040)
-0.530***
(0.0050)
A. Only positive expenditure on transport fuels
B. Include all household size
C. Original data set without data cleaning
-0.507***
(0.0130)
-0.518***
(0.0210)
-0.508***
(0.0040)
-0.535***
(0.0060)
Appendix – A Carbon tax case (1/2)
29
Estimated carbon tax revenues from household
sector totaled 8.92 Billion Baht
Consumption
Billion
groups
Baht
Electricity
1.96
Fuels for transport
2.47
Food and beverages
1.66
Others
2.84
Note
From price increase by 17.15%
From price increase by 9.21%
From price increase by 1.27%
From price increase by 1.99%
Appendix – A Carbon tax case (2/2)
30

Middle-income households suffer the largest welfare losses from
energy consumption (direct effect)

Indirect effect on non-energy consumption is regressive, indicating
relatively larger burden on low-income households
Mean monthly
consumption per
household (Baht)
Direct effect
Indirect effect
Total effect
Quintile 1
7,087
1.15%
1.42%
2.57%
Quintile 2
10,227
1.25%
1.41%
2.66%
Quintile 3
13,416
1.32%
1.40%
2.72%
Quintile 4
18,456
1.31%
1.38%
2.69%
Quintile 5
34,749
1.30%
1.29%
2.59%
All
16,787
1.27%
1.38%
2.65%
Equivalized
consumption
Relative welfare losses*
*Relative welfare losses are measured by CV as a percentage of consumption