ABSTRACT
Title of dissertation: ESSAYS ON ENVIRONMENTAL ECONOMICS:
CARBON TAX, PRICE REGULATION AND
RESIDENTIAL ELECTRICITY CONSUMPTION
Jikun Wang
Doctor of Philosophy, 2019
Dissertation directed by: Professor Roberton Williams III
Department of Agricultural and Resource Economics
This dissertation features three essays in environmental economics. In the
first essay, I analyze how price regulation changes the welfare effects of carbon
emission policies. Specifically this paper shows that electricity price regulation in
China substantially increases the cost of reducing carbon emissions. I set up a simple
general equilibrium model where the price of the carbon intensive sector (electricity)
is regulated and examine the welfare impact of a revenue-neutral carbon tax. The
model shows price regulation has a direct cost effect and also changes the secondary
cost effects caused by pre-existing taxes. I then construct a static CGE model where
the parameters are calibrated using stylized facts about the economy of China in
2007. My central estimate shows the marginal cost of achieving a 20% reduction in
CO2 emissions is 22% higher in the presence of pre-existing taxes than in a world
with only a carbon tax. Price regulation raises the marginal cost of CO2 reduction
by a further 27%, on top of the distortion caused by pre-existing taxes.
The second essay studies the implication of relaxing electricity price regulation,
both in the context of pre-existing taxes and in the context of carbon pricing policy.
It employs a general equilibrium model of the Chinese economy and provides ex-
ante counterfactuals under a range of electricity regulation policy and assess the
social welfare impact with potential electricity market reform. It shows pre-existing
labor tax increases the per unit social benefit of deregulation. The analysis also
shows carbon emission policy increases the per unit benefit of electricity deregulation
compared with second-best setting.
The third essay uses micro-level data in China to study the impact of urban
heat island on residential electricity use (REU). Combining a household energy
consumption survey in China with remote sensing data related to urban heat island
intensity and weather conditions specific to each household, the empirical analysis
shows that urban heat island has a significant effect on residential electricity use,
through interacting with local weather conditions such as temperature. Higher
urban heat island increases residential electricity consumption by increasing the
impact of Degree Days on REU. The effect also varies seasonally and regionally.
ESSAYS ON ENVIRONMENTAL ECONOMICS:
CARBON TAX, PRICE REGULATION AND RESIDENTIAL
ELECTRICITY CONSUMPTION
by
Jikun Wang
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2019
Advisory Committee:
Professor Roberton Williams III, Chair/Advisor
Professor Anna Alberini
Professor Maureen Cropper
Professor Joshua Linn
Professor James Archsmith
?c Copyright by
Jikun Wang
2019

Dedication
To My Loved Ones.
ii
Acknowledgments
A doctoral dissertation is the culmination of the PhD experience, while the journey
itself, with its ebbs and flows, is much more than that. I wish to thank everybody
who support me along the way from the very beginning.
First of all, I thank my advisor Roberton Williams III for his invaluable sup-
port, advice and guidance. I learned a great deal about being an economist and being
a researcher in general from his insight, intellectual curiosity and passion. I greatly
appreciate his encouragement when I hit a rough patch during the PhD program. I
absolutely enjoy our conversations on all sorts of things related to economics, which
is one of the things I will miss the most.
I would also like to thank Anna Alberini, Maureen Cropper, Joshua Linn and
James Archsmith. I am grateful to have them serving on my dissertation committee.
They provided me with so many insightful discussions and constructive comments
throughout the development of my dissertation.
During my time at Maryland, I had the opportunity to get to know and work
with many other faculty members, including Lint Barrage, Robert Chambers, Snae-
bjorn Gunnsteinsson, Jim Hanson, Sebastien Houde, Jorge Holzer, Richard Just,
Ted McConnel and Lars Olson. I learned so much from being in their classes, being
in workshops or simply talking with them. They helped me grow as a researcher
and as a person.
I would also like to thank many of the wonderful people in AREC who helped
to make my life as a graduate student a lot easier, including Jeff Cunningham,
Katherine Faulkner, Liesl Koch and Pin-Wuan Lin.
iii
At Maryland, I had the opportunity to meet a terrific group of fellow doctoral
candidates, many of which became dear friends. The struggles of graduate school
are diminished when shared with those who have experienced the same. I want to
thank in particular Xing Hong, Qiong Peng, Hong Lin, Tzu-yao Lin, Ziyan Yang,
Yuan Shi, Ying Zeng, Xiru Zhang, Pan He, Li Fang, Xu Han, Aaron Adalja, Mehrab
Bakhtiar, Davide Cerruti, Jaclyn Evans, Jeff Ferris, Patrick Fleming, Joe Maher,
Mark Miller, Guanghui Que, Dan Werner, Youpei Yan, Jun Zhang and Christy
Zhou.
My cohort at the University of Maryland deserve a special mention. Charlene
Chi, Jen Zheng He, Min Kim, Yuangdong Qi, Casey Wichman and Jane Zhou were
excellent companions during my years in the program. I miss fighting in the trenches
together during the first year of our program and I miss spending time with them.
Last but not least, I would like to deeply thank the loved ones in my life for
their unconditional love and support. I thank my parents for their endless support,
not only during my graduate study, but throughout my life. I thank Ivy for being
a constant source of happiness in my life and for being there for me. I couldnt have
done this with without you.
iv
Table of Contents
Preface ii
Foreword ii
Dedication ii
Acknowledgements iii
Table of Contents v
List of Tables vii
List of Figures viii
1 Electricity Price Regulation and the Cost of Reducing Carbon Emission 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 What?s unique about China . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Electricity prices and Marginal cost of production . . . . . . . 6
1.3 The Analytical Model . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Baseline Scenario Welfare Analysis . . . . . . . . . . . . . . . 10
1.3.3 Price Regulation and its effects on Welfare Analysis . . . . . . 11
1.4 The Numerical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.1 Model Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4.1.1 Firms . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.1.2 Household . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.1.3 Government Policy . . . . . . . . . . . . . . . . . . . 18
1.4.1.4 Equilibrium Conditions . . . . . . . . . . . . . . . . 19
1.4.2 Data and Parameters . . . . . . . . . . . . . . . . . . . . . . . 19
1.4.3 Proposed Policy . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.5 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.5.1 Comparing with the First Best . . . . . . . . . . . . . . . . . 23
1.5.2 Comparing with the Second Best . . . . . . . . . . . . . . . . 25
1.5.3 Toal Welfare Impacts from Optimal carbon tax and Pigouvian
tax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.6 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.7 Conclusion and Caveats . . . . . . . . . . . . . . . . . . . . . . . . . 34
v
2 Welfare Analysis of Deregulation in Chinese Electricity Sector 36
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.1 Model Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.1.1 Firms . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.1.2 Household . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3.1.3 Government Policy . . . . . . . . . . . . . . . . . . . 44
2.3.1.4 Equilibrium Conditions . . . . . . . . . . . . . . . . 45
2.3.2 Data and Parameters . . . . . . . . . . . . . . . . . . . . . . . 45
2.3.3 Carbon Emission Policy . . . . . . . . . . . . . . . . . . . . . 48
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.4.1 Central case . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.4.2 Comparison between Second best and First Best . . . . . . . . 51
2.4.3 Considering Productivity Gains from Deregulation . . . . . . . 53
2.4.4 Carbon Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5 Conclusion and Caveats . . . . . . . . . . . . . . . . . . . . . . . . . 58
3 Residential Energy Use and Heat Island Effect 61
3.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.1 MODIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.2 CRECS 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.3 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.4.1 Overall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.4.2 Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.4.3 Winter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.4.4 North vs South . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.4.5 Additional Analysis . . . . . . . . . . . . . . . . . . . . . . . . 85
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
A Chapter 1 89
A.1 A: Analytical Derivations . . . . . . . . . . . . . . . . . . . . . . . . . 89
A.1.1 deriving equation 1.6 . . . . . . . . . . . . . . . . . . . . . . . 89
A.1.2 deriving equation 1.7 . . . . . . . . . . . . . . . . . . . . . . . 91
A.1.3 deriving equation 1.11 . . . . . . . . . . . . . . . . . . . . . . 91
A.1.4 deriving equation 1.13 . . . . . . . . . . . . . . . . . . . . . . 93
A.2 B: the Numerical Model . . . . . . . . . . . . . . . . . . . . . . . . . 94
A.2.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
A.2.2 Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Bibliography 96
vi
List of Tables
1.1 Benchmark Data for Numerical Model . . . . . . . . . . . . . . . . . 21
1.2 Percentage Increases in Marginal Cost at Different Levels of Emission
Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.3 Marginal Abatement Cost under Alternate Parameter Values and
Model Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.1 Data for Model Calibration . . . . . . . . . . . . . . . . . . . . . . . 47
3.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3 Dwelling Szie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.4 Built Year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.5 Urban . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.6 Ownership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7 Regression Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.8 Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.9 Winter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.10 North . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.11 South . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.12 Single Family Home and Uran Home . . . . . . . . . . . . . . . . . . 86
3.13 Built after 1990 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
vii
List of Figures
1.1 China experimental carbon trading zone . . . . . . . . . . . . . . . . 5
1.2 China?s Electricity prices and Coal prices . . . . . . . . . . . . . . . . 7
1.3 Marginal cost of emissions reduction . . . . . . . . . . . . . . . . . . 24
1.4 Marginal cost of emissions reduction . . . . . . . . . . . . . . . . . . 25
1.5 Net Welfare Gain from Optimal level of regulation . . . . . . . . . . 29
1.6 Net Welfare Gain under the Pigouvian rule . . . . . . . . . . . . . . . 30
1.7 Production elasticities sensitivity analysis . . . . . . . . . . . . . . . . 31
1.8 Utility function elasticities sensitivity analysis . . . . . . . . . . . . . 31
1.9 Initial tax rate sensitivity analysis . . . . . . . . . . . . . . . . . . . . 32
1.10 Price of electricity sensitivity analysis . . . . . . . . . . . . . . . . . . 33
2.1 Welfare Impacts of Electricity Price Regulation . . . . . . . . . . . . 50
2.2 Effects of Pre-existing Taxes . . . . . . . . . . . . . . . . . . . . . . . 53
2.3 Higher Initial Labor Tax Rate . . . . . . . . . . . . . . . . . . . . . . 54
2.4 With Productivity Gains . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5 With Carbon Emission Reduction Target . . . . . . . . . . . . . . . . 57
3.1 Box plot of CDD and HDD by month . . . . . . . . . . . . . . . . . . 65
3.2 Land Cover Type legend and class definitions . . . . . . . . . . . . . 66
3.3 Households in CRECS 2012 . . . . . . . . . . . . . . . . . . . . . . . 68
3.4 Profiles of household characteristics comparing CRECS2012 and NBS2013
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
viii
Chapter 1: Electricity Price Regulation and the Cost of Reducing
Carbon Emission
1.1 Introduction
Climate change is one of the most pressing issues in this century and the key
to solving it is to reduce carbon emission. Emission pricing as a market-based policy
has been proven to be an efficient method. When studying the effect of emission
pricing on economy, a free market is generally assumed. One result of a free market
policy is that an emission fee will most likely increase electricity prices and people
will respond by reducing demand for electricity. However, some degree of price
regulation always exists, so an emission fee will translate into less or no increase
in the price of electricity. This is particularly important in China, the number one
carbon emitter in the world and where price regulation in the electricity sector is
strict. Consequently, a key question remains unanswered: how will price regulation
change the effect of emission pricing on the economy?
Carbon emission regulation is a classic example of environmental regulation.
The problem of optimal environmental regulation arises from the fact that environ-
mental amenities are public goods. Environmental amenities will cause externality
1
in the system, either on the consumer side, which is affecting the utility function
of agents (like clean air, greater visibility), or on the production side, which is af-
fecting the production function (such as the effect of climate change on agricultural
production).
In a world with lump-sum transfers, the optimal environmental regulation
needs to equate the marginal benefit of changing environmental goods to the marginal
cost. In the case of taxation, the optimal tax should be equal to the Pigouvian tax,
which is equal to the marginal social benefit from the environmental improvement
stemming from a one unit reduction in the polluting good. The Pigouvian rule is an
application of the Samuelson condition for optimal level of public good provision.
In a second best world without lump-sum transfers, the optimal environmen-
tal taxation is generally a combination of a distortionary (Ramsey) component and
an environmental (Pigouvian) component [Goulder, 1995, Bovenberg and Goulder,
1996, Goulder et al., 1999, Bovenberg and Goulder, 2002] . The Ramsey part repre-
sents the distortion caused by distortionary taxes for raising public funding to finance
public goods without lump-sum transfers. The environmental (non-distortionary)
part tries to capture the externality caused by polluting goods, as discussed above.
The distortionary taxes in the system are of first-order importance in determin-
ing the optimal environmental taxation. Also for the same reason, to evaluate the
welfare impact of a revenue neutral environmental tax reform, one must take into
account the pre-existing distortions in the system.
In a model with one representative agent (income coming only from labor
input) and a competitive market with CRS production technology, Bovenberg and
2
van der Ploeg [1994] have shown that the optimal tax rate on dirty intermediate
goods is not just equal to the Pigouvian tax, but also divided by the Marginal Cost
of Public Funding (MCPF), representing the wedge created by distortionary taxes.
The tax rate on dirty final goods and the tax rate on wages have a relationship as
discussed above.
For the special case of weakly separable environmental/public goods, weakly
separable leisure, and homothetic consumption of clean and dirty goods in the util-
ity function, Bovenberg and Mooij [1994] and Bovenberg and Goulder [1996] also
show that the welfare change of a revenue neutral environmental tax reform would
generally have a negative gross cost caused by a decrease of labor supply and erosion
of the tax base. Thus, a double dividend is not the case. But the overall welfare
impact may well be positive after taking the environmental benefits into account
[Parry et al., 1999, Bento and Jacobsen, 2007].
Later papers have considered the general externality of environmental goods
by relaxing the assumptions about environmental goods being weakly separable in
the utility function [Williams III, 2002, Carbone and Smith, 2008, Fullerton and
Kim, 2008, Cremer et al., 2010] . The basic intuition follows. The gross cost stems
from the tax interaction effect.
This paper extends the literature by introducing a third constraint: certain
industries might be regulated in terms of price. This is particularly relevant for the
carbon trading market in China. In several pilot cities where carbon trading exists,
the majority of the carbon emission comes from the electricity generating sector,
which is heavily regulated in China.
3
Generally speaking, higher output prices mean lower relative wages and lower
private returns on labor. This leads to less labor supply and exacerbates the tax
interaction effect by increasing the deadweight loss of distortionary labor taxes. So
policies that will have a lower output price effect will be less distortionary. One
example is energy efficiency standards [Goulder et al., 2016] . Intuitively, the same
logic holds for regulated sector prices. (Depending on how profits are modeled, there
might be additional income effects as well.)
I set up a general equilibrium model of an economy with two production sectors
(dirty and clean) and price regulation in the dirty sector. I look at the welfare impact
of a revenue-neutral tax reform when the output price of the polluting good is fixed.
The model analytically examines the tradeoff between the tax interaction effect and
the revenue recycling effect. On one hand, the tax burden does not fully pass onto
households in the form of higher consumer demand prices. On the other hand,
consumer income decreases due to less profit from the energy sector.
1.2 What?s unique about China
As the world?s number one carbon emitter, China has launched several experi-
mental carbon trading zones (city level or provincial level, all currently separate , see
Figure 1.1) in an effort toward a national carbon trading scheme. If built, it would
be the world?s largest carbon emission trading system and the most important in
the world?s battle with climate change [Freeman and Kong, 2013] .
China?s carbon emission reduction tends to be associated with reductions in
4
Figure 1.1: China experimental carbon trading zone
air pollutants such as sulfur dioxide and particulate matter because of the nation?s
energy dependency on coal. So carbon emission trading could have the added benefit
of improving air quality.
While China has pledged to cut the carbon intensity of its economy?as mea-
sured by the amount of carbon emitted per unit of GDP?by 40%-45% from 2005
levels by 2020, local Chinese governments have little incentive to participate due to
economic development concerns and the global scope of GHG damages . However,
concerns about air pollution makes its reduction a political priority. Also, when
a future national trading scheme is set up, the central government would want to
consider associated different air quality benefits in allocating the emission reduction
efficiently across locations.
Another feature that is particular to the case of China is the government?s
control of markets. Despite the major shift from a central planning economy to
5
market based economy starting from the late 1970s, there are still sectors that are
run mostly by the state owned enterprises and controlled by the government, such as
the energy sector. The Chinese state still retains a tight grip on oil product prices,
natural gas prices, and electricity tariffs. For example, the national development and
reform commission (NDRC) sets the price schedules for residential and industrial
electricity periodically. This market rigidity prevents energy firms in China from
passing on cost increases to consumers.
1.2.1 Electricity prices and Marginal cost of production
Since coal is the primary component of marginal cost of electricity generation
in China, a quick back of envelop calculation using price of coal (350 RMB/ton 1)
and heat rate of coal generation (370 grams of coal equivalent per kWh from Zhang
et al. [2017b] ) gives us a rough estimate of marginal cost of electricity production,
about 0.13 Yuan/kWh. The average electricity price is 0.52 Yuan/kWh 2.
Another way of looking at the relative relationship between electricity price
and marginal cost of production is look at the overall trend of electricity price and
coal prices. Figure 1.2 (Source: Liu et al. [2013]) shows the price index of electricity
and coal.
1https://www.ceicdata.com/en/china/price-monitoring-center-ndrc-market-and-contract-
price-coal?page=2
2https://www.ceicdata.com/en/china/electricity-price/cn-service-price-36-city-avg-electricity-
for-resident-220v
6
Figure 1.2: China?s Electricity prices and Coal prices
1.3 The Analytical Model
Here I present an analytical model to compare the gross costs of environmental
policies in the presence of distortionary taxes and price regulation. Subsection 1 lays
out the model assumptions and set-up. Subsection 2 shows the general equilibrium
welfare impacts of a revenue neutral carbon tax policy. Subsection 3 introduces
price regulation into the model and reexamines how price regulation changes the
gross costs of the same carbon tax policy.
7
1.3.1 Model Assumptions
A representative agent allocates its time endowment (L?) between labor supply
(L) and leisure (l = L??L). It also consumes two goods, X and Y . Y represents an
aggregate of final output from industries that use fossil fuels (D) intensively, and X
is an aggregate of all other consumption goods. There?s no capital accumulation in
the economy, so the model focuses on behavior in one period, rather than solving a
dynamic problem.
The household tries to maximize
U = u(X, Y, l)? ?(D) (1.1)
where u(?) is continuous and quasi-concave. ? is disutility from current, an-
thropogenically caused additions to the stock of CO2 in the atmosphere, caused
by combustion of fossil fuels. I assume that the use of fossil fuel input D leads
to a proportional amount of carbon emissions and choose the units to imply one
ton of carbon per unit of D. This is a standard assumption which says there?s no
economically viable scrubber technologies to reduce CO2 emissions per unit of fuel
input. The separability in (1.1) implies that future climatic change does not alter
the relative decisions between leisure and two types of consumption.
Final goods are produced by firm with constant returns to scale technology
using clean (nonpolluting) input C and dirty input D. The production function are
8
given by
X = X(Cx, Dx) (1.2)
Y = Y (Cy, Dy) (1.3)
Intermediate goods C,D are produced with labor only, and the marginal prod-
uct of labor in each of these two industries is taken to be constant. All four industries
are assumed to be competitive.
The use of dirty input leads to a proportional amount of carbon emissions. I
choose units to imply one ton of carbon per unit of D.
Finally, I assume the government spends G as lump-sum transfers to house-
holds. The government raises revenue by levying a proportional tax tL on labor
income and carbon emission tax ?D. The government budget is assumed to balance,
with
G = tLL+ ?DD
Denoting the demand prices of X and Y by Px and Py and normalizing the
gross wage to unity, I can write the household budget constraint as
PxX + PyY ? (1? ?L)L+G (1.4)
From the first-order conditions and household budget constraint, I obtain the
(implicit) uncompensated demand and labor supply functions.
Finally, the equilibrium level for fossil fuel use D(?D, tL) and labor L(?D, tL)
will be functions of exogenous parameters ?D and tL.
9
1.3.2 Baseline Scenario Welfare Analysis
First, I consider the case where price regulation doesn?t exist. Consider a
policy change involving a shift in ?D while holding government transfer G constant
in nominal terms. The government budget is assumed to balance, and therefore any
revenue consequences from regulations are neutralized by adjusting the rate of tL.
The welfare effect of the policy change is obtained by differentiating the utility
function with respect to ?D, allowing tL to vary. This gives
dU ?u dX ?u dY ?u dl dD
= + + ? ?? (1.5)
d?D ?X d?D ?Y d?D ?l d?D d?D
Substituting consumer?s first order conditions, firms? first order conditions ,
and total differentiation of production function and resource constrain into Eq(1.5
) 3
1 dU ?
= (?D ?
? dD dL
) + tL (1.6)
? d?D ? d?D d?D
This expression clearly states the source and magnitude of welfare conse-
quences of emission abatement. The welfare effect (in dollars) can be separated
?
into two term. The first part is distortion from carbon emission, where ?D ? ? is?
the wedge between marginal efficiency cost of dirty input and marginal environmen-
tal cost of carbon emission. If the tax on dirty input is set at Pigouvian tax level
?
(?D =
? ), then the distortion from carbon emission will be zero. The second part
?
is distortion from labor market, which is the wedge between the net wage (equal to
3see appendix for details
10
the marginal social cost of labor in terms of foregone leisure) and the gross wage
(equal to the value of marginal product of labor), multiplied by the change in labor
supply.
The second part is what previous literature of preexisting taxes (especially
distorted labor markets) explores. Because of the distortion in labor market, pres-
ence of preexisting taxes significantly raises the general equilibrium costs of carbon
policies. This part could be further broken into two effects, tax-interaction effect
and revenue-reclying effect.
Combining the labor demand function and government budget constraint, the
equation above becomes 4
1 dU ?? dD dD ?L
= ?(?D ??? ) ? + Z? (D +???D ?) + (?Z + 1?)?tL (1.7)? d?D ? d?D d?D ?tD?
dW p ?WR ?W I
where
?t ?LL ?t
Z = L (1.8)
L+ t ?LL ?tL
1.3.3 Price Regulation and its effects on Welfare Analysis
Suppose carbon policy is implemented in an economy where the price of carbon
intensive good Y regulated. Here I consider the case where price regulation is strict,
i.e. Py is fixed at a certain level. Therefore Py becomes a parameter (later I set
Py = P?y ). The difference between price Py and true cost of producing Y means
the firm incurs potential profit (or loss), I. Firms are assumed to be owned by the
4see appendix for details
11
government.
The difference between this case and the baseline scenario lies mainly at the
firm side. The firm producing Y solves
max?[qCCy + (qD + ?D)Dy]
s.t.
Y (Cy, Dy) ? Y
Profit I of the firm is given by
I = PyY ? qCCy ? (qD + ?D)Dy (1.9)
The new government budget is
G = I + tLL+ ?DD (1.10)
Similarly, I derive the comparative statics following a tax shift on dirty input
5
1 dU ? ?
? dD dY dL
= (?D ) + (Py ? ?) + tL (1.11)
? d?D ? d?D d?D d?D
Here I introduce a new parameter ?. ? is the Lagrangian multiplier from the
firm?s cost minimizing problem and in equilibrium it equals the marginal cost to the
5see appendix for details
12
firm for producing Y . When the price of Y is exogenously determined, Py (which is
equal to the marginal social benefit) is no longer necessarily equal to the marginal
cost of Y , so their difference (Py ? ?) represents the distortion in Y sector.
I further break down the last term in Equation 1.11. I get a similar expression
containing tax-interaction terms and revenue-recycling terms.6
1 dU ??? dD ? dY dI dD ?L ?L dPy= (?D ) +(Py ?) +Z( +D+?D )+(Z+1)tL +(Z+1)tL
? d?D ? d?D d?D d?D d?D ?tD ?Py d?D
(1.12)
When Py = P?y, dPy = 0. Therefore, the expression above becomes
1 dU ??? dD ? dY dI dD ?L= (?D ) +(Py ?) +Z( +D+?D )+(Z+1)tL (1.13)
? d?D ? d?D d?D d?D d?D ?tD
Comparing with the expression where price is not regulated, there are two
major changes. First, (P ??) dYy term is added to the primary cost part. This termsd?D
measures the additional primary welfare impact that is caused by the distortion
in market of Y , which is equal to the wedge between marginal social benefit and
marginal social cost of producing Y , times the change in Y caused by emission tax.
For example, if the price of Y is set above the marginal cost of producing Y , then it
is welfare improving to increase the Y ?s production and welfare reducing to decrease
its production. Change in welfare depends on how much Y changes as a response
to the change in ?D. Hence, the welfare impact is the product of those two parts.
6see appendix for details
13
If the price is set at the competitive price level in the absence of price regulation,
this term equals zero for incremental changes in ?d. However, for large increases in
?d, the marginal cost of producing Y for the firm will rise, and the whole term will
represent a negative welfare effect.
The second change is the third term Z( dI +D + ? dDD ) which represent thed?D d?D
tax-recycling effect. It still measures the efficiency gain from the using government
revenues to finance cuts in distortionary taxes. This is the product of marginal
excess burden of taxation and the marginal revenue. The marginal revenue now
is the sum of marginal revenue from the emission tax and marginal profit of firms
which is transferred to the government.
1.4 The Numerical Model
The analytical model in section 3 is great for providing a clear intuition for the
mechanisms, but less for gauging the real world importance of the price regulation of
interest. In this section, I extend the original model to better analyze the magnitude
of this issue.
First, the numerical model allows me to consider the effects of ?large?policy
changes. For example, I could now look at carbon tax that produces greater than
incremental emissions reduction and evaluate its overall welfare impact rather than
marginal effect. To do this, I must specify functional forms for the utility and
production functions and solve the model computationally. The second extension
is to disaggregate further the group of goods in production and make every good
14
both input and output during production. This allows for a richer model of the real
economy and a richer sensitivity analysis. I also fit the numerical model to a real
sized economy.
Subsection 1 lays out the model structure and describes the behavior of house-
holds, firms and the government, and the equilibrium conditions of the numerical
model. It also assumes their functional forms. A more detailed description of the
model is in the appendix. Subsection 2 describes the calibration of the model.
1.4.1 Model Structure
There are four categories of goods. Fossil fuels (F ) represent the sector that is
under carbon emission control. Electricity (E) is the sector that is price regulated
in this model. The rest of goods in the economy is grouped into an aggregate of
goods that are non-energy-intensive (CN) and an aggregate of goods that are energy
intensive (CI). The electricity industry and fossil fuels industry are directly affected
by the policies considered in the paper and they are the key of my analysis, therefore
I want to keep them as two distinctive sectors. I also want to have at least two other
sectors to represent the rest of the goods such that households would enjoy some
extent of substitution among other goods.
As in the analytical model, there is no capital accumulation and intertemporal
investment. Labor (L) is the sole primary factor of production and is equal to the
household time endowment net of leisure. All the goods are used as inputs for the
production of other goods and are consumed by the household.
15
As in the analytical model, labor is taxed at a proportional rate of ?L and
carbon tax is at a per unit rate of ?t. There is an additional consumption tax
(?C which is proportional to consumer goods prices. The government collects tax
revenue to finance in real terms a fixed level of lump-sum transfers to households
and public goods provision .
The extended model is solved numerically to obtain exact welfare implications,
rather than marginal welfare effects obtained above. This is useful for evaluating
?large?reductions in emissions.
1.4.1.1 Firms
I assume competitive producers that take input and output prices as given. I
use a constant elasticity of substitution (CES) form for production functions in all
sectors: ?
?
X = a ( ? X j
?
+ ? L j)1/?jj j ij ij Lj j (1.14)
i
where i = {CI , CN , CF , CE}, j = {CI , CN , CF , CE} , the ??s and ??s are pa-
rameters, ? = (? ? 1)/? and ? is the elasticity of substitution between factors in
production. Because this production function possesses constant returns to scale,
supply curves in all sectors are perfectly elastic for given input prices.
Firms choose input quantities to maximize profits (or minimize costs of pro-
duction). Profits equal the value of output minus expenditures on labor and inter-
mediate goods used in production, less any charges per unit of carbon emissions.
16
?
?j = (pj ? bj?t)Xj ? piXij (1.15)
i
where pi and pj are the prices of inputs and outputs, and bj is emission intensity
factor that equals the CO2 emissions per unit of good j. bj is zero for all goods
except fossil fuels (F ).
Note that when output price pj is endogenously determined, because the pro-
duction function exhibits constant returns to scale, in equilibrium profits will equal
zero under carbon tax. When output price is regulated and pj is exogenously de-
termined, the firm will incur some profit or loss depending on whether price is set
at a level higher or lower than the marginal cost of producing Xj. In that case,
since I assume firms (under price regulation) are state owned, remaining profits are
transferred back to the government.
1.4.1.2 Household
I assume the representative household has a CES utility function as following:
?
U = U(CI , CN , CF , CE, l, e) = ( ? C
?u + ? l?u)1/?ui i l ? ?(e) (1.16)
i
where l is leisure time and i = {CI , CN , CF , CE} , the ??s and ?u are param-
eters, ?u = (?u ? 1)/?u and ?u is the elasticity of substitution between leisure and
consumption goods. e denotes carbon emissions.
17
The household maximizes utility subject to the budget constraint:
?
(1 + ?c)piCi = pL(1? ?L)L+ pcGH (1.17)
i
where ?c is the tax rate on consumption and ?L is the tax rate on labor income,
L = T ? l is labor supply, GH is government transfers to households in real terms,
pc is consumer goods price index. In the following analysis, taxes finance a fixed
level of government transfers to households.
1.4.1.3 Government Policy
The numerical model considers the same type emission regulation considered
in the analytical model: a tax of ?t on carbon emissions. A carbon permit or emission
quota policy can be readily implemented in the model to generate the same level of
abatement and gives the same welfare implications assuming permits are auctioned
off and all that revenue is used by the government to finance reductions in preexisting
tax rates (firms behave identically under the carbon tax and under the quota for a
given level of emission reduction).
The government?s budget constraint is given by:
? ?
piGi + pcGH = ?LpLL+ ?c piCi + ?te (1.18)
i i
18
With price regulation for electricity sector, the government?s budget constraint is :
? ? ?
piGi + pcGH = ?LpLL+ ?c piCi + ?te+ ?i (1.19)
i i
where Gi is government spending in each sector, GH is government transfers
to households (all fixed in real terms), ?i is firm?s profit. The tax rate ?L is adjusted
to compensate for the effects of emission policy on government revenue.
1.4.1.4 Equilibrium Conditions
In equilibrium, supply of each good equals the demand which is the sum of
household and government consumption. Labor supply which is total time endow-
ment net of leisure equals sum of labor demand across all sectors. Government bud-
get constraint is met by adjusting labor tax ?L to neutralize tax revenue changes
from emission policy. And total carbon emission equals the target level.
1.4.2 Data and Parameters
The model benchmark data set, which is summarized in Table 1.1, approxi-
mates China?s economy in 2007. I start with a micro level social account matrix of
China?s economy with more than 60 industries and aggregate it up to 4 sectors. The
carbon emission data is from World Bank world development indicators dataset. I
use it to calculate the carbon intensity parameter for fossil fuels. 7
The distribution parameters for production functions are calibrated based on
7Assume there is no viable end-of-pipe carbon emission reduction technology, so carbon emission
from fossil fuel consumption is proportional to fossil fuel output
19
the assumed elasticities of substitution and the identifying restriction that each
industry is cost minimizing over mix of inputs, or equivalently, the restriction that
in the absence of a new emission reduction policy, the model will replicate the
benchmark data.
An important preference parameter is ?u, the consumption-leisure substitu-
tion elasticity. I choose this, along with the labor time endowment, to imply an
uncompensated and compensated labor supply elasticity of 0.15 and 0.3 respec-
tively. These are typical estimates from the literature. The pre-existing labor tax
and consumption tax is estimated from benchmark data to be 0.18 and 0.05 re-
spectfully. I vary these parameters in the sensitivity analysis later to test to what
extent my simulation results depend on the parameters specification. Again, the
distribution parameters of utility production are calibrated based on the assumed
elasticities of substitution and the identifying restriction that household is utility
maximizing subject to budget constraint, or equivalently, the restriction that in the
absence of a new emission reduction policy, the model will replicate the benchmark
data.
20
21
Table 1.1: Benchmark Data for Numerical Model
CN CI CF CE labor household consumption tax labor tax government total
CN 38,054.9 59,483.8 1,933.3 2,828 66,472 31,410 200,182.6
CI 52,755.8 265,206.8 5,141.1 8,361 152,065 11,092 494,621.9
CF 969.0 13,475.6 852.7 5,781 726 21,804.1
CE 4,163 18,671 1,961 1,967 1,699 28,460.4
labor 94,105 116,084 9,764 7,540 227,493.4
household 227,493 5,446 232,939.4
consumption tax 11,978 11,978.0
labor tax 10,135 21,700 2,152 1,983 35,970.2
government 11,978 35,970 47,948.2
total 200,182.6 494,621.9 21,804.1 28,460.4 227,493.4 232,939.4 11,978.0 35,970.2 47,948.2
1.4.3 Proposed Policy
The Paris Agreement calls for nations to achieve significant reductions in green-
house gases (including CO2) through ?nationally determined contributions? (NDCs)
to hold the increase in the global average temperature to well below 2 degrees Celsius
above pre-industrial levels and to pursue efforts to limit the temperature increase to
1.5 degrees Celsius above pre-industrial levels. China has since pledged to achieve
the peaking of carbon dioxide emissions around 2030 and increase the share of non-
fossil fuels in primary energy consumption to around 20 percent by 2030. In my
numerical simulations, I consider reducing carbon emissions from the fossil fuels
sector ranging from 0 % to 25 %, a range that is relevant to the reductions needed
to fulfill China?s commitment.
I consider a range of estimates for benefits from carbon abatement. Central
estimates in the literature for social cost of carbon (SCC) are typically around 30-
40 US dollars per ton of CO 82 . Converting to 2007 levels considering the inflation
and exchange rates gives us about 500-600 Renminbi per ton of carbon. Of course,
this represents a global marginal benefit from carbon abatement. If considering
the regional SCC, that is when only the damages to that particular region are
included in the calculation, only a few IAMs disaggregate the global SCC. Nordhaus
[2017] provides a estimates that China?s regional SCC is about 20 % of the global
level. Considering the very large uncertainties about the magnitude of the SCC, the
8Revised estimates from Nordhause [Nordhaus, 2017] is 31$ per ton of CO2 in 2010 US $ for
2015. The U.S. Interagency Working Group on Social Cost of Greenhouse Gases (2016) [Working
Group on Social Cost of Carbon, 2016] estimates around $40 per ton of CO2
22
simulations below span a wide range of benefit scenarios, from 0 to 900 Renminbi
per ton of carbon. In all cases, I assume that marginal benefits are constant over
the range of emissions reductions.
1.5 Numerical Results
In this section, I present the main results from solving the numerical model and
policy counterfactuals. I emphasize the relative costs of carbon policy under different
settings, as opposed to absolute costs. Together with sensitivity analysis, I examine
how costs are affected by parameters of production function, utility function, levels
of pre-existing taxes and levels of pre-determined electricity price (for the cases
applicable). These parameters affect cost by determining the relative contributions
of input-substitution and output-substitution effect and the degree to which price
regulation and distortionary taxes interact.
1.5.1 Comparing with the First Best
First I start from the first best. Figure 1.3 depicts the marginal cost of reducing
CO2 emissions at different percentage reduction levels for the cases with and without
electricity price regulation. There is no pre-existing labor or consumption taxes
associated with either of the two cases. The marginal cost is substantially higher
when the electricity price is regulated.
The difference represents the primary welfare effect of introducing price regu-
lation into the economy. Electricity price regulation imposes an additional welfare
23
800
price regulation with no pre-existing taxes
no price regulation & no pre-existing taxes
700
600
500
400
300
200
100
0
0 5 10 15 20 25
Percentage Reduction in Carbon Emissions
Figure 1.3: Marginal cost of emissions reduction
cost even when distortionary labor and consumption taxes do not exist. This cor-
roborates my findings from the analytical model. When the electricity price is fixed
at the competitive price level in absence of price regulation, the additional welfare
cost term starts from zero. Therefore the marginal cost curve under price regulation
also has a zero intercept. The additional cost increases at a faster pace as more
reduction in carbon emissions are realized ( which means a higher emission tax and
higher marginal cost of producing electricity). In the graph, the gap between those
two curves expands as we move towards more reduction in carbon emissions.
24
Marginal Cost(2007 Chinese yuan/ton)
1.5.2 Comparing with the Second Best
The result above serves as a reference point. Of course in the real economy,
there are plenty of pre-existing distortionary taxes and it is very unlikely that price
regulation is the only distortion. Therefore, it is important to look at a case where
price regulation is introduced on top of pre-existing taxes and examine its welfare
implication.
900
price regulation & pre-existing taxes
only pre-existing taxes
800 none
700
600
500
400
300
200
100
0
0 5 10 15 20 25
Percentage Reduction in Carbon Emissions
Figure 1.4: Marginal cost of emissions reduction
Figure 1.4 shows how the presence of price regulation, interacting with distor-
tionary taxes, significantly raises the costs of reductions in carbon emissions for the
proposed carbon tax policy.
25
Marginal Cost(2007 Chinese yuan/ton)
The bottom curve is the marginal cost of reductions in CO2 emissions (2007
Chinese yuan/ton of CO2), when there is no distortionary labor tax or consumption
tax. The curve is upward sloping, reflecting the increasing difficulty of substituting
fossil fuels for other inputs in production.
In a second-best world, the marginal cost curve (represented by the middle
curve in Figure 1.4 ) is raised significantly, proving the importance of the revenue-
recycling effect and the tax-interaction effect that are induced by pre-existing dis-
tortionary taxes. Also, similar to previous studies, under carbon tax policy, the
marginal cost curve pivots upward but retains the zero intercept that applies in the
first-best case. The zero intercept reflects that the revenue-recycling effect exactly
offsets the tax-interaction effect at the first increment of abatement.
The top curve in Figure 1.4 represents the marginal cost reductions in CO2
emissions, when the electricity sector price is regulated. This curve is upward sloping
and also has a zero intercept. The zero intercept reflects the fact that when the
electricity sector price is fixed at the competitive price level in absence of price
regulation, the direct distortionary effect from price regulation starts from zero,
while the revenue-recycling effect exactly offsets the tax-interaction effect at the
first increment of abatement.
Comparing the three curves, the curve with price regulation sits on top of
both the other two curves throughout the entire range of carbon emission reduc-
tion. This shows that the presence of price regulation, interacting with distortionary
taxes, substantially raises the costs of reductions in carbon emissions. With price
regulation, welfare cost for the same level of emission tax is lower, while emission re-
26
duction is also lower . Overall for the same level of emission reduction, the marginal
cost per unit of emission is higher compared with the case without price regulation.
It is also worth noting that to achieve the same level of emissions reduction
requires a higher carbon tax rate under the case with price regulation.
Table 1.2: Percentage Increases in Marginal Cost at Different Levels of Emission
Reduction
percentage reduction
in CO2 emission
Percentage Increase in Marginal Cost 5% 10% 15% 20%
pre-existing taxes & no price regulation 18.5 19.9 20.0 21.8
vs no pre-existing taxes & no price regu-
lation
pre-existing taxes & price regulation vs 12.7 16.2 21.3 26.6
pre-existing taxes & no price regulation
pre-existing taxes & price regulation vs
no pre-existing taxes & no price regula- 33.5 39.4 45.6 54.1
tion
Table 1.2 shows the increases in marginal costs when comparing the three
cases discussed above. Distortionary taxes on average raise the cost by around 20%
compared to the first best. When price regulation is introduced on top of that,
marginal cost is increased by a similar magnitude. It is also worth noting that the
discrepancy in costs rises at a much faster rate when price regulation is present,
compared with when there is only pre-existing taxes. This shows that from an
efficiency perspective, taking price regulation into consideration is just as, if not
more important than considering the effects of distortionary taxes when evaluating
and choosing the best environmental policies.
27
1.5.3 Toal Welfare Impacts from Optimal carbon tax and Pigouvian
tax
Total welfare impact estimate shows us the magnitude of net benefits from
the carbon policy and the effects of price regulation relative to that of pre-existing
taxes. The optimal emissions reduction is easily inferred from Figure 1.4. It is where
a given (constant) marginal benefits curve intersects the applicable marginal cost
curve. Figure 1.5 shows the maximum welfare gain? that is, the welfare gain from
setting emissions reduction as discussed before? under each setting as a function
of marginal environmental damages per ton of carbon emissions. As shown, for any
level of damages, the maximum welfare gain under second best setting is lower than
when there is no labor tax. Price regulation will lower that furthermore.
Figure 1.6 presents the welfare impact under each setting, if the Pigouvian or
first best rule is followed: that is, if carbon tax is set equal to the SCC.
1.6 Sensitivity Analysis
Table 1.3 presents the sensitivity analysis as I vary the parameter values and
model specifications. I vary the elasticity of substitution in production, elasticity of
substitution in consumption, the pre-existing labor tax rate and the level at which
electricity price is set when price regulation is present. These parameters determine
the relative contributions of input-substitution and output-substitution effect and
the degree to which price regulation and distortionary taxes interact. I look at the
28
1000
none
only pre-existing taxes
900 price regulation & pre-existing taxes
800
700
600
500
400
300
200
100
0
0 100 200 300 400 500 600 700
Marginal damages (2007 yuan/ton of carbon)
Figure 1.5: Net Welfare Gain from Optimal level of regulation
how the model behaves when there is only pre-existing taxes but no price regulation
and compare that with the case where both pre-existing taxes and price regulation
are present.
In the second row, I vary the elasticity of substitution in production. A higher
elasticity of substitution in production means a lower primary cost of reducing emis-
sions, which will lower the overall marginal cost. Therefore as I increase the elasticity
of substitution, marginal costs of both cases (no price regulation and price regula-
tion) are lower. But the relative efficiency of the two cases does not change since this
mainly affects the primary marginal cost part for both in the same way. Figure 1.7
29
Net benefit (billions of 2007 yuan per year)
1200
none
only pre-existing taxes
price regulation & pre-existing taxes
1000
800
600
400
200
0
0 100 200 300 400 500 600 700 800 900
Marginal damages (2007 yuan/ton of carbon)
Figure 1.6: Net Welfare Gain under the Pigouvian rule
shows the detailed result.
In the third row, I vary compensated and uncompensated labor supply elas-
ticity which results in a change in the elasticity of substitution parameter of the
utility function. The results are mainly driven by this parameter. ?u = 0.00, ?c =
0.15 implies an elasticity of substitution around 1, and ?u = 0.30, ?c = 0.45 implies
an elasticity of substitution around 3, compared to around 2 in the central case.
In my model, the elasticity of substitution for the household measures the
degree of substitution between consumption and leisure as well as between different
consumption goods. Making it easier for consumers to choose goods that are less
30
Net benefit (billions of 2007 yuan per year)
600 250
price regulation & pre-existing taxes price regulation & pre-existing taxes
only pre-existing taxes only pre-existing taxes
500
200
400
150
300
100
200
50
100
0 0
0 5 10 15 20 25 0 5 10 15 20 25 30
Percentage Reduction in Carbon Emissions Percentage Reduction in Carbon Emissions
(a) 3/2 of base elasticity (b) 4 times base elasticity
Figure 1.7: Production elasticities sensitivity analysis
carbon intensive and consumes less carbon intensive goods in response to a carbon
tax increase will lower the primary marginal cost of reducing carbon emissions. At
the same time, higher elasticities of substitution between consumption and leisure
strengthens the tax-interaction and revenue-recycling effects and will increase costs
of reducing carbon emissions. Overall the primary cost component dominates the
increased tax-interaction and revenue-recycling effects. (See Figure 1.8 for details).
800 900
price regulation & pre-existing taxes price regulation & pre-existing taxes
only pre-existing taxes only pre-existing taxes
800
700
700
600
600
500
500
400
400
300
300
200
200
100
100
0 0
0 2 4 6 8 10 12 14 16 18 20 0 5 10 15 20 25
Percentage Reduction in Carbon Emissions Percentage Reduction in Carbon Emissions
(a) ?u = 0.00,?c = 0.15 (b) ?u = 0.30, ?c = 0.45
Figure 1.8: Utility function elasticities sensitivity analysis
31
Marginal Cost(2007 Chinese yuan/ton) Marginal Cost(2007 Chinese yuan/ton)
Marginal Cost(2007 Chinese yuan/ton) Marginal Cost(2007 Chinese yuan/ton)
In the forth row, I vary the pre-existing labor tax rate and consumption tax
rate. A higher distortionary tax will result in higher marginal costs across the
board. For a higher initial tax rate, the distortion on the labor market is greater.
It enhances both the revenue recycling and tax-interaction effect. Again it does not
change the relative efficiency of the two cases (See Figure 1.9).
800 1000
price regulation & pre-existing taxes price regulation & pre-existing taxes
only pre-existing taxes only pre-existing taxes
900
700
800
600
700
500
600
400 500
400
300
300
200
200
100
100
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Percentage Reduction in Carbon Emissions Percentage Reduction in Carbon Emissions
(a) ?L = 0.0, ?c = 0.0 (b) ?L = 0.3, ?c = 0.05
Figure 1.9: Initial tax rate sensitivity analysis
Last but not least, I vary the level at which the price of electricity is set
when it is under price regulation. This only affects the second case when both
price regulation and pre-existing taxes are present. As expected, setting the price
of electricity further apart from its competitive price level when there is only pre-
existing taxes enhances the distortion in the electricity sector and increases the cost
of reducing carbon emissions. Also, the distortion intensifies faster for higher P?y
as more emissions reduction is realized. This is reflected in Figure 1.10. The gap
between those two curves expands fasters for higher P?y as we move towards more
reduction in carbon emissions.
32
Marginal Cost(2007 Chinese yuan/ton)
Marginal Cost(2007 Chinese yuan/ton)
700 700
price regulation & pre-existing taxes price regulation & pre-existing taxes
only pre-existing taxes only pre-existing taxes
600 600
500 500
400 400
300 300
200 200
100 100
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Percentage Reduction in Carbon Emissions Percentage Reduction in Carbon Emissions
(a) P?y = 1.2 (b) P?y = 1.5
Figure 1.10: Price of electricity sensitivity analysis
Table 1.3: Marginal Abatement Cost under Alternate Parameter Values and Model
Specifications
without price regulation with price regulation
5% 10% 20% 5% 10% 20%
1. Central Case
75.3 179.5 492.9 84.9 208.6 623.9
2. Production elasticities
3/2 of base elasticity 53.2 122.0 316.3 59.6 141.2 374.7
4 times of base elasticity 21.1 46.9 109.9 24.3 53.2 122.8
3. Utility function elasticities
?u = 0.00, ?c = 0.15 88.3 203.7 564.9 97.7 235.5 702.8
?u = 0.30, ?c = 0.45 68.1 166.2 452.5 77.9 195.9 584.6
4. Initial tax rate
?L = 0.0, ?c = 0.0 63.6 149.7 404.8 73.9 174.8 497.8
?L = 0.3, ?c = 0.05 84.3 199.9 535.2 98.4 231.8 669.0
5. Price of electricity
P?y = 0.9 75.3 179.5 492.9 84.9 209.2 625.3
P?y = 1.2 75.3 179.5 492.9 89.7 215.7 629.6
P?y = 1.5 75.3 179.5 492.9 97.1 223.0 646.6
33
Marginal Cost(2007 Chinese yuan/ton)
Marginal Cost(2007 Chinese yuan/ton)
1.7 Conclusion and Caveats
In this paper, I have employed analytical and numerical general equilibrium
models to examine how price regulation changes the efficiency impacts of revenue-
neutral carbon taxes in a second-best setting with pre-existing labor taxes. For the
policies considered, the efficiency costs are considerably higher than would be the
case in the absence of price regulation.
The effect of price regulation is present even when there are no distortionary
taxes in the system. Price regulation alone considerably raises the cost of reducing
carbon emission. This reflects the fact that price regulation brings a direct distortion
to the economy.
This distortion continues to affect the cost and effectiveness of carbon tax
policy in a second-best setting. For the same level of carbon tax, the reduction
in emissions achieved when there is price regulation is lower. While for the same
level of reduction in carbon emissions, the marginal cost per unit is higher when
the price is regulated. This relationship holds across a wide range of parameter
values and model specifications. The higher cost reflects interaction between price
regulation and pre-existing distortions stemming from labor and consumption taxes.
As a price is controlled, it lessens the distortions created by pre-existing taxes, but
it also dampens the response that a proposed policy aims for. Overall it drives up
the efficiency cost.
The associated efficiency disadvantage can be very large: my central estimate
is that, in the presence of price regulation, marginal cost of reduction is generally
34
raised by 27% to achieve a 20% reduction in carbon emissions, compared with the
case without, and raised by 54% when compared with the first best.
Of course there are some limitations to my present study. First, the ana-
lytical and numerical models are static. I abstract from the dynamics of capital
accumulation and economic growth. I also ignore the impacts of carbon policy on
investments in research and technology and the associated potential to reduce future
costs of abatement. These dynamic effects may have significant welfare consequences
that are not covered by my analysis.
Second, this analysis treats the benefit of reducing carbon emissions as con-
stant and additively separable. This partially relates to the static nature of my
model. Similar to the dynamics of reducing the cost of abatement in the future,
the dynamic issues associated with the benefits from carbon abatement are absent
from my analysis. Also, modeling benefits as additively separable helps in deriv-
ing a clean analytical expression of welfare impacts and makes it easier in adding
the welfare gains from carbon abatement policies to consider their full welfare im-
pacts. However, this might understate the prospects for welfare gains since there
are ancillary benefits from carbon abatement such as benefits from the reduction in
other fossil-fuel related pollutants (for example, fine particulate matter, especially
for the case of China) and these benefits tend to interact with labor supply decisions,
people?s health and productivity.
35
Chapter 2: Welfare Analysis of Deregulation in Chinese Electricity
Sector
2.1 Introduction
The first chapter has examined the implications of electricity price regulation
for the cost of reducing carbon emission. It substantially increases the welfare cost
of carbon pricing policy to achieve the same level of CO2 emission reduction and
therefore substantially increases the cost per unit of carbon reduction. It takes the
electricity price regulation as given. It is worthwhile to explore the potential of
relaxing that restriction. The chapter studies the implication of relaxing electricity
price regulation, both in the context of pre-existing taxes and in the context of
carbon pricing policy.
Over the past three decades, many member countries of the OECD and more
than 70 developing countries have taken steps to reform their electricity sectors.
[Bacon and Besant-Jones, 2001, Khanna and Rao, 2009]
Much of this literature (see summary by Jamasb [2005], Besant-Jones [2006],
Khanna and Rao [2009] ) uses cross-country panel data or plant-level data to con-
trol for within-country variations in regional and/or plant-level characteristics to
36
estimate the impact of reform on performance in the electricity sector.
These studies indicate that reforms in electricity sector improved generation
efficiency, labor productivity, capacity utilization, and supply quality while reducing
frequency of blackouts and electricity losses and led to gains in social welfare. Re-
forms also improved the fiscal position of governments by approximately 1 percent
of GDP (Malik et al. [2015], Pombo and Taborda [2006], Besant-Jones [2006], Zhang
et al. [2008])
This paper supplements the literature that studies the reform and deregulation
of electricity sectors with a quantitative CGE model aimed to provide counterfactual
ex-ante comparisons and assessing the economy-wide welfare impact of potential
electricity market reform.
One key issue in examining the efficiency impacts of deregulating the electric-
ity sector is pre-existing distortions in the economy such as income and consump-
tion taxes. Literature in environmental taxation has provided good argument and
evidence that pre-existing taxes will amplify the distortions (in the form of new
environmental taxes) introduced into the economy. Studies on trade find the same
effect. Potential gains or losses from tariffs, import or export restrictions are magni-
fied through interaction with pre-existing taxes [Williams III, 1999]. Following the
logic, it could mean that electricity price deregulation, which reduces distortions,
will yield additional efficiency gains. Few papers that assess the welfare gain from
price de-regulation uses the general equilibrium approach and takes into account
pre-existing taxes.
This study provides insights into how much welfare gain is not accounted
37
for by missing the general equilibrium effects, particularly from pre-existing taxes.
My numerical model indicates that a noticeable difference in welfare gain from
deregulating electricity price when comparing the simulation results under first best
setting and the second best setting.
This paper also contributes to the literature of China?s effort to reduce CO2
emissions and its interaction with electricity sector reform 1. The most closely
related studies include Goulder and Morgenstern [2017] and Teng et al. [2017]. They
look at the impact of electricity market reform and deregulation on the performance
of electricity sector and the effectiveness of carbon policy. What differs my analysis
from theirs is the adoption of a general equilibrium model of the entire economy
instead of a partial equilibrium model of the electricity sector. By modeling the
effects of deregulation of electricity prices on prices in other sectors, as well as the
income effect on households, my model allows the demand of electricity to adjust
in equilibrium. My model gives a comprehensive analysis of the economy wide
impacts of electricity deregulation and allows for interaction between electricity
price regulation and carbon pricing policy in the economy.
Specifically, for a given CO2 emission target (e.g. 6 percent), the model solves
for the optimal carbon pricing depending on the electricity price regulation level and
assess the welfare benefit from deregulation. The analysis finds that deregulating
electricity price still brings substantial social welfare benefit, holding the emission
reduction target constant. It also shows that the per unit benefit of electricity
1A number of recent empirical studies on China?s effort to reduce CO2 emissions through
emission trading include Ho et al. [2017], Teng et al. [2017], Zhang et al. [2017a]
38
deregulation is higher compared with second-best and the benefit is higher when
the emission reduction target is set higher.
The rest of this chapter is organized as as follows. Section 2 provides the
background on Chinese electricity sector and its relevance to this study. Section 3
describes the model used in numerical analysis. In section 4, I discuss the results.
Section 5 concludes.
2.2 Background
Since the ?open and reform? in 1978, China has largely gone from a centrally
planned economy to a market-based economy. Prices are mostly determined by the
market place. The power(electricity) sector remains one of the few major exceptions.
Transmission and distribution fees as well as wholesale and retail electricity prices
are set by the NDRC, a government regulating body. Since 1985, the electricity
sector has been going through a long and slow process of reforming. There has
been plans for many years to de-regulate this sector. The goal is to introduce
competition in both wholesale and retail markets, and gradually allow prices to be
more responsive to supply and demand [Ho et al., 2017].
The current electricity sector in China features a highly regulated dispatch
and pricing system. For pricing, the wholesale and retail electricity prices are set
by the government.
For dispatch, there is not a spot market for electricity. Dispatch is largely
done by the economic planning commissions at the provincial government level,
39
using a ?equal share dispatch? principle. In this dispatch system, the provincial
dispatch authorities first allocate annual generation quotas to individual generation
companies, based on annual demand forecast and equal annual utilization hours
for a given technology type (e.g. 600 MW coal fire units). The dispatching center
then breaks them down to quarterly, monthly, and daily unit generation schedule.
Transmission and distribution fees are set by the government as well.
This paper focuses on the deregulation of electricity pricing, which brings
wholesale and retail prices more in line with production cost.
By abstracting from the deregulation in dispatch (from an equal share alloca-
tion to a more cost-effective approach), this analysis misses the potential efficiency
gains from competition which drives down production cost. In terms of emission
reduction, deregulation in dispatch incentivize switching to less carbon intensive
generation technology and renewable energy. This brings the added benefit of ac-
celerating transitioning the economy to sustainable energy source.
Deregulation in the electricity sector is also related to Chinas plan to fight
climate change. China is one its way to build a national emission trading system
(ETS) [Goulder et al., 2017] . Once fully built, it will double the size of CO2 emission
covered by some form of carbon pricing globally. The power sector is an especially
critical sector in terms of the nations climate policy effort. It accounts for over 40
percent of the countrys total CO2 emissions Yang and Lin [2016] . It is the first
sector to be covered by the national ETS. From the perspective of designing and
implementing a good and effective carbon pricing policy, its important to examine
on current institution of the electricity generation and transportation and prepare
40
for potential changes.
2.3 Model
2.3.1 Model Structure
The model builds on the one from the first chapter. There are four categories
of goods. Fossil fuels (F ) represent the sector that is under carbon emission control.
Electricity (E) is the sector that is price regulated in this model. The rest of goods
in the economy is grouped into an aggregate of goods that are non-energy-intensive
(CN) and an aggregate of goods that are energy intensive (CI). The electricity
industry and fossil fuels industry are directly affected by the policies considered in
the paper and they are the key of my analysis, therefore I want to keep them as
two distinctive sectors. I also want to have at least two other sectors to represent
the rest of the goods such that households would enjoy some extent of substitution
among other goods.
As in the analytical model, there is no capital accumulation and intertemporal
investment. Labor (L) is the sole primary factor of production and is equal to the
household time endowment net of leisure. All the goods are used as inputs for the
production of other goods and are consumed by the household.
As in the analytical model, labor is taxed at a proportional rate of ?L and
carbon tax is at a per unit rate of ?t. There is an additional consumption tax
(?C which is proportional to consumer goods prices. The government collects tax
revenue to finance in real terms a fixed level of lump-sum transfers to households
41
and public goods provision .
The extended model is solved numerically to obtain exact welfare implications,
rather than marginal welfare effects obtained above. This is useful for evaluating
?large?reductions in emissions.
2.3.1.1 Firms
I assume competitive producers that take input and output prices as given. I
use a constant elasticity of substitution (CES) form for production functions in all
sectors: ?
? ?
Xj = a
j j 1/?j
j( ?ijXij + ?LjLj ) (2.1)
i
where i = {CI , CN , CF , CE}, j = {CI , CN , CF , CE} , the ??s and ??s are pa-
rameters, ? = (? ? 1)/? and ? is the elasticity of substitution between factors in
production. Because this production function possesses constant returns to scale,
supply curves in all sectors are perfectly elastic for given input prices.
Firms choose input quantities to maximize profits (or minimize costs of pro-
duction). Profits equal the value of output minus expenditures on labor and inter-
mediate goods used in production, less any charges per unit of carbon emissions.
?
?j = (pj ? bj?t)Xj ? piXij (2.2)
i
where pi and pj are the prices of inputs and outputs, and bj is emission intensity
factor that equals the CO2 emissions per unit of good j. bj is zero for all goods
except fossil fuels (F ).
42
Note that when output price pj is endogenously determined, because the pro-
duction function exhibits constant returns to scale, in equilibrium profits will equal
zero under carbon tax. When output price is regulated and pj is exogenously de-
termined, the firm will incur some profit or loss depending on whether price is set
at a level higher or lower than the marginal cost of producing Xj. In that case,
since I assume firms (under price regulation) are state owned, remaining profits are
transferred back to the government.
2.3.1.2 Household
I assume the representative household has a CES utility function as following:
?
U = U(CI , CN , CF , CE, l, e) = ( ?iC
?u
i + ?ll
?u)1/?u ? ?(e) (2.3)
i
where l is leisure time and i = {CI , CN , CF , CE} , the ??s and ?u are param-
eters, ?u = (?u ? 1)/?u and ?u is the elasticity of substitution between leisure and
consumption goods. e denotes carbon emissions.
The household maximizes utility subject to the budget constraint:
?
(1 + ?c)piCi = pL(1? ?L)L+ pcGH (2.4)
i
where ?c is the tax rate on consumption and ?L is the tax rate on labor income,
L = T ? l is labor supply, GH is government transfers to households in real terms,
pc is consumer goods price index. In the following analysis, taxes finance a fixed
43
level of government transfers to households.
2.3.1.3 Government Policy
The numerical model considers the same type emission regulation considered
in the analytical model: a tax of ?t on carbon emissions. A carbon permit or emission
quota policy can be readily implemented in the model to generate the same level of
abatement and gives the same welfare implications assuming permits are auctioned
off and all that revenue is used by the government to finance reductions in preexisting
tax rates (firms behave identically under the carbon tax and under the quota for a
given level of emission reduction).
The government?s budget constraint is given by:
? ?
piGi + pcGH = ?LpLL+ ?c piCi + ?te (2.5)
i i
With price regulation for electricity sector, the government?s budget constraint is :
? ? ?
piGi + pcGH = ?LpLL+ ?c piCi + ?te+ ?i (2.6)
i i
where Gi is government spending in each sector, GH is government transfers
to households (all fixed in real terms), ?i is firm?s profit. The tax rate ?L is adjusted
to compensate for the effects of emission policy on government revenue.
44
2.3.1.4 Equilibrium Conditions
In equilibrium, supply of each good equals the demand which is the sum of
household and government consumption. Labor supply which is total time endow-
ment net of leisure equals sum of labor demand across all sectors. Government bud-
get constraint is met by adjusting labor tax ?L to neutralize tax revenue changes
from emission policy. And total carbon emission equals the target level.
2.3.2 Data and Parameters
The model benchmark data set, which is summarized in Table 2.1, approxi-
mates China?s economy in 2007. I start with a micro level social account matrix of
China?s economy with more than 60 industries and aggregate it up to 4 sectors. The
carbon emission data is from World Bank world development indicators dataset. I
use it to calculate the carbon intensity parameter for fossil fuels. 2
The distribution parameters for production functions are calibrated based on
the assumed elasticities of substitution and the identifying restriction that each
industry is cost minimizing over mix of inputs, or equivalently, the restriction that
in the absence of a new emission reduction policy, the model will replicate the
benchmark data.
An important preference parameter is ?u, the consumption-leisure substitu-
tion elasticity. I choose this, along with the labor time endowment, to imply an
uncompensated and compensated labor supply elasticity of 0.15 and 0.3 respec-
2Assume there is no viable end-of-pipe carbon emission reduction technology, so carbon emission
from fossil fuel consumption is proportional to fossil fuel output
45
tively. These are typical estimates from the literature. The pre-existing labor tax
and consumption tax is estimated from benchmark data to be 0.14and 0.05 respect-
fully. I vary these parameters in the sensitivity analysis later to test to what extent
my simulation results depend on the parameters specification. Again, the distribu-
tion parameters of utility production are calibrated based on the assumed elasticities
of substitution and the identifying restriction that household is utility maximizing
subject to budget constraint, or equivalently, the restriction that in the absence of
a new emission reduction policy, the model will replicate the benchmark data.
46
47
Table 2.1: Data for Model Calibration
CN CI CF CE labor household consumption tax labor tax government total
CN 38,054.9 59,483.8 1,933.3 2,828 66,472 31,410 200,182.6
CI 52,755.8 265,206.8 5,141.1 8,361 152,065 11,092 494,621.9
CF 969.0 13,475.6 852.7 5,781 726 21,804.1
CE 4,163 18,671 1,961 1,967 1,699 28,460.4
labor 94,105 116,084 9,764 7,540 227,493.4
household 227,493 5,446 232,939.4
consumption tax 11,978 11,978.0
labor tax 10,135 21,700 2,152 1,983 35,970.2
government 11,978 35,970 47,948.2
total 200,182.6 494,621.9 21,804.1 28,460.4 227,493.4 232,939.4 11,978.0 35,970.2 47,948.2
2.3.3 Carbon Emission Policy
The Paris Agreement calls for nations to achieve significant reductions in green-
house gases (including CO2) through ?nationally determined contributions? (NDCs)
to hold the increase in the global average temperature to well below 2 degrees Celsius
above pre-industrial levels and to pursue efforts to limit the temperature increase to
1.5 degrees Celsius above pre-industrial levels. China has since pledged to achieve
the peaking of carbon dioxide emissions around 2030 and increase the share of non-
fossil fuels in primary energy consumption to around 20 percent by 2030. In my
numerical simulations, I consider reducing carbon emissions from the fossil fuels
sector by 6% and 13 %, a range that is relevant to the reductions needed to fulfill
China?s commitment.
2.4 Results
I consider a range for electricity price regulation stringency. Electricity price
is expressed as a proportion of the marginal cost of production. Therefore, the
extent to which electricity price differs from its marginal production cost reflect the
stringency of price regulation. When equal to its marginal cost, electricity price
is essentially determined by the market and is considered the no regulation case.
I consider electricity price regulation ranging from 50 percent of its marginal cost
of production to two times the marginal cost. I calculate the social welfare cost
when moving the price away from marginal production cost, in reference to the no
regulation case.
48
With an empirical estimate of the ratio between electricity price and marginal
cost of production, one could locate the current state of the electricity sector along
the curve. The social welfare cost associated with that point could then be inter-
preted as the welfare benefit of deregulation electricity price from the current state
to being completely market-determined.
I simulate the same policy change under three settings. The central case as-
sumes the actual labor and consumption tax rates from data and the model replicates
the baseline economy. Since there are existing distortions in the economy, I refer to
this as the second-best setting in the discussion below. The second one represents an
economy without distortionary taxes. I take the calibrated parameters from central
case, set labor tax and consumption tax to zero and run the model to simulate such
an economy. I refer to this as the first-best setting below. The last one represents
an economy with an emission reduction target. I assume carbon tax is implemented
to achieve such target. This is referred to as the carbon pricing setting.
2.4.1 Central case
Figure 2.1 shows the economy wide welfare cost under different levels of price
regulation in the electricity sector, all compared with the no regulation case. The
U-shaped curve demonstrate two findings. First, there is positive cost when price is
regulated above or below marginal cost. That confirms the analysis from classical
economic theory.
The welfare cost of price regulation consists manly a dead weight loss. When-
49
350
300
250
200
150
100
50
0
0.5 1 1.5 2
Electricity Price / Marginal Production Cost
Figure 2.1: Welfare Impacts of Electricity Price Regulation
ever there exists a wedge between the marginal benefit (represented by the price)
and marginal cost of economic activity, it creates a dead weight loss. Second, as we
move away from the middle along the X axis, the cost is increasing quadratically.
This reflects the fact that the size of dead weight loss is proportional to the wedge
of prices as well as change in quantity in that sector. Both are increased as we move
the ratio of electricity price and marginal production cost away from one.
The magnitude of welfare cost is also northing noting. The electricity sec-
tor is worth about 2.846 trillion Yuan. The numerical results show the benefit of
deregulating the electricity market is substantial.
50
 Cost(billion 2007 Chinese yuan)
2.4.2 Comparison between Second best and First Best
A key advantage of the generally equilibrium approach is that the model cap-
tures the interaction of policy reform with pre-existing distortions in the economy
and assesses the welfare implication from that interaction.
To illustrate the importance of second-best consideration, I calculated the
welfare cost under first-best setting and compare them to the results from second-
best.
To facilitate the comparison of results under different settings, I adjust the
absolute level of welfare cost by the changes in electricity production level. The
reason is two folds. First, when the model is re-calibrated to the first-best setting,
the reference point is different. The baseline economy under first best setting is
larger and the electricity production level is higher. Since the deadweight loss is
generally proportional to the size of economy, it calls for welfare cost measures that
account for that change. The adjustment used here calculates something similar to
the average cost instead of total cost of the electricity price regulation.
Another reason for the adjustment is that it makes sure the comparison is not
confounded by the change in the primary welfare component of welfare cost. In the
addition the primary welfare cost which is the deadweight loss from electricity sector,
the total social welfare cost under the second-best setting captures the effect of pre-
existing taxes. As discussed above, the size of the first part is proportional to the
size of distortion in electricity sector as well as the changes in electricity production
level. By making the adjustment, the difference in cost measures between first-best
51
and second-best setting could be seen as the welfare implication from interacting
with pre-existing distortions.
Generally speaking, the addition welfare impact from pre-existing taxes is a
combination of the policy interaction effect and revenue recycling effect. For ex-
ample, when electricity price is set above marginal production cost, it raises the
equilibrium prices and overall price level, which reduces labor supply. This exacer-
bates the distortions from labor market and is the policy interaction effect. At the
same time, since generators have profits and it could be recycled to reduce labor
tax rate, it reduces the labor sector distortions and is called the revenue-recycling
effect. The net effect is line with the primary welfare effect.
Figure 2.2 shows the difference of social welfare cost per unit of electricity.
It shows that under the second-best scenario, pre-existing distortions magnify the
welfare cost of electricity price regulation.
This mechanism holds when electricity is priced above and below marginal
cost. When electricity is priced above marginal cost, it is similar to having a tax
and when electricity is priced blow marginal cost, it is similar to putting a subsidy
on electricity. Previous literature often looks at taxes (or policy instruments with
the same effect as a tax) in numerical analysis and this provides an example that
the magnifying effect holds true for policies that acts as a subsidy.
Another thing worth noting is that the additional welfare cost represented by
the gap between curves is proportional to the primary effect. It squares with findings
in the literature. This also ensures the cost curve under second-best reaches zero
when electricity price is equal to its marginal production cost.
52
0.5
First Best
Second Best
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0.5 1 1.5 2
Electricity Price / Marginal Production Cost
Figure 2.2: Effects of Pre-existing Taxes
Perhaps whats surprising about the numerical results is that the magnitude
of difference between second and first best setting is small. This could be partially
explained by the low level of initial labor tax rate in the simulation. This is confirmed
by the sensitivity analysis when adopting a high initial labor tax rate of 0.45. See
Figure 2.3
2.4.3 Considering Productivity Gains from Deregulation
While this paper focuses on the economy wide welfare gains of deregulation
from correctly aligning cost and benefit, empirical studies indicate that reforms in
53
Welfare Cost(2007 Chinese yuan/kWh)
0.8
First Best
Second Best
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.5 1 1.5 2
Electricity Price / Marginal Production Cost
Figure 2.3: Higher Initial Labor Tax Rate
electricity sector also improved industry productivity such as generation efficiency,
labor productivity, capacity utilization. Production efficiency gains could also come
from the deregulation in dispatching (from an equal share allocation to a more
cost-effective approach).
By altering the parameter of productivity of the electricity generation industry,
the numerical model provides a way of examining the significance of productivity
gains. Teng et al. [2017] suggests an overall 10 percent improvement in generation
efficiency when China moves to a deregulated spot market and dispatching system.
Following their estimates, Figure 2.4 presents the comparison between the central
54
Welfare Cost(2007 Chinese yuan/kWh)
case and the second best effect of productivity gains. 3
0.9
First Best
Second Best
0.8 Second Best with 10%production efficiency gain
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.5 1 1.5 2
Electricity Price / Marginal Production Cost
Figure 2.4: With Productivity Gains
2.4.4 Carbon Pricing
The third setting in which I run the quantitative analysis assume a exogenous
carbon emission reduction target (6 percent and 13 percent). I assume the economy
adopts efficient carbon pricing such as a carbon tax to achieve such goal. I hold
the reduction level constant across electricity regulation levels and let the model
solve for optimal level of carbon tax. I use the no regulation case as reference point
3The analysis assumes gradual increase of the productivity parameter for electricity sector, as
the price/MC ratio moves toward 1
55
Welfare Cost(2007 Chinese yuan/kWh)
to calculate the social welfare cost of moving electricity price away from market
determined level.
Of course, electricity price regulation creates incentive to increase or decrease
electricity production. Since electricity production is a major contributor of carbon
emission, electricity price regulation affects the total carbon emission level, as well
as the carbon tax needed to achieve the same emission target. The numerical results
from the model shows as price moves up with respect to marginal cost, the optimal
carbon tax decrease.
Figure 2.5 shows the results under 6 percent carbon emission reduction target
and 13 percent reduction target and compares the cost per unit of electricity across
all three settings.
Studies that look at the interaction of Chines electricity market reform and
carbon emission trading system find the deregulation in electricity market would
achieve higher emission reduction target or reduce the cost of carbon pricing for
achieving the same emission target, when carbon tax or ETS is the policy instru-
ment. This study assess the welfare impact of relaxing price regulation itself.
When deregulating electricity price, especially if the original level is above
marginal cost, it means a higher carbon tax or tighter allowance in ETS is needed
to reach the same level of carbon emission reduction. This generally means increased
welfare cost from carbon pricing and if it dominates the welfare benefit from dereg-
ulation, it could reverse the total welfare impact of deregulating electricity price.
By holding the reduction carbon constant, and allowing for optimal carbon tax to
be endogenously determined in equilibrium, my analysis presents the total welfare
56
0.6
First Best
Second Best
Carbon Reduction 1
Carbon Reduction 2
0.5
0.4
0.3
0.2
0.1
0
0.5 1 1.5 2
Electricity Price / Marginal Production Cost
Figure 2.5: With Carbon Emission Reduction Target
impact of deregulation in the carbon pricing setting.
Of course, the total welfare measure here doesnt include the environmental
benefit that carbon tax brings to the society. However, since I hold the CO2 emission
constant across board, the environmental benefit from mitigated climate change
should be the same. Therefore, one could compare the welfare cost across different
level of electricity price regulation.
Figure 2.5 also compares the cost per unit of electricity across settings. The
welfare cost is larger under carbon reduction setting, compared to the second best
and first best. And its large when carbon reduction level is higher, which requires
57
Welfare Cost(2007 Chinese yuan/kWh)
a higher-level carbon tax across board. This represents the potential welfare im-
provement from electricity price deregulation is also higher in the context of carbon
reduction, at least when its implemented in an economically efficient way such as
carbon tax or ETS.
The rationale behind this result is similar to before. The new carbon tax is
another distortion introduced in the system and its welfare effect is magnified when
electricity price moves further away from efficient level due to regulation.
2.5 Conclusion and Caveats
China?s plan for electricity sector deregulation bolds important implications
for social welfare and carbon emission policy.
This chapter employs a general equilibrium model of the Chinese economy and
provides ex-ante counterfactuals under a range of electricity regulation policy and
assess the social welfare impact with potential electricity market reform.
It exploits the substitution in consumption and production of electricity as
a results of such price deregulation and allows for aggregate electricity demand to
adjust in equilibrium. The numerical results shows with moderate assumption on
the stringency of current electricity price regulation, the market reform could bring
substantial welfare benefit.
The generally equilibrium approach also allows for considering the effects of
pre-existing distortions. As shown by the numerical simulations, pre-existing labor
tax increase the benefit per unit of electricity when considering the same degree of
58
electricity price deregulation. This magnifying effect is closely related to the level
of pre-existing tax rate. The higher the labor tax rate, the larger the magnifying
effect. With the current labor tax rate implied by the data, the difference between
first best and second best setting is small.
This chapter also applies the model to look at the welfare impact of electricity
market reform in the context of carbon emission policy. Electricity price regulation
affects the effectiveness and cost of carbon emission policy. In my model, it also
implies the level of carbon tax needed to achieve certain emission level is affected by
the extent of electricity price regulation. The analysis shows deregulating electricity
price will still bring substantial social welfare benefit, holding the emission reduction
target constant. The model also shows assuming a 6 percent and 13 percent carbon
emission reduction and efficient carbon pricing as the policy instrument, the per
unit benefit of electricity deregulation is higher compared with second-best. And
the benefit is higher when higher emission reduction target is set.
The study has several limitations. First, without a detailed modeling of the
structure of the electricity sector, the study misses the potential efficiency gains
from competition which drives down production cost.
The stylized model also abstracts from any heterogeneity among household
and firms. The analysis thus misses the distributional effect of electricity sector
deregulation, which is equally important as social welfare effects when designing
and implementing policy.
When considering the interaction between deregulation and carbon emission
policy in China, the analysis assume an economic efficient policy such as carbon
59
tax or cap-and-trade is used, when in reality China is on its way to build its ETS
with tradable performance standard (TPS). As evidenced by Goulder et al, the cost-
effectiveness of TPS is different from a cap-and-trade counterpart, and its interaction
with potential deregulation is different as well.
60
Chapter 3: Residential Energy Use and Heat Island Effect
3.1 Motivation
As a major component of electricity consumption, Residential Electricity Use
(REU) contributes significantly to the pressure of climate change. In 2017, residen-
tial electricity use accounts for about 30 percent of global electricity use1.
Studies have shown that both social economic and climatic factors affect how
much electricity a household uses [Wolfram et al., 2012, Auffhammer and Mansur,
2014] . Social economic determinants include income, family size, home size , and
etc. Climatic/meteorological factors include local temperature, humidity, and etc
[Davis and Gertler, 2015]. Generally speaking, hot days will increase people?s cool-
ing needs and cold days will increase people?s heating needs. Both lead to more
electricity consumption. One could expect with the trends in climate change, we
will see more extreme weather conditions, which in return will increase REU.
REU is on the rise with fast urbanization in most of the developing world.
Urbanization is associated with changes in many of the social economic factor dis-
cussed above. It means changed life style, increased income and growing urban
population. For instance, a typical urban household in China consumes 1.4 times
1http://www.eia.gov/tools/faqs/faq.cfm?id=447&t=1
61
as much energy as its rural counterpart, with heating and cooling accounting for
more than 50 percent of the total REU [Zheng et al., 2014]. As the global urban
population is expected to increase by another 3 billion, mostly from Africa and Asia
(86 percent) [Heilig, 2012], curbing the soaring REU is crucial.
With urbanization, there has also been an increasing interest on the role of
urban spatial forms on electricity consumption [Seto and Dhakal, 2014]. A compact
development pattern, marked by high residential density and land use mix, leads to
lower REU[Ewing and Rong, 2008, Jones and Kammen, 2014, Ahmad et al., 2015]. A
compact building environment decreases REU demand and increase energy efficiency
[Ewing and Rong, 2008, Lee and Lee, 2014].
A missing piece, however, is how the urban heat island effect (UHI), another
byproduct of urbanization, affects REU. The agglomeration of population and in-
frastructure raises the urban temperature through higher anthropogenic heat emis-
sion, lower evapotranspiration and larger building volume [Oke, 1982, Yow, 2007,
Stone Jr and Rodgers, 2001]. Thus, UHI could potentially affect summer cooling
and reducing winter heating.
One way UHI could affect REU is that through interacting with local weather
conditions to change their marginal impact on residential electricity use. Presum-
ably, places with higher UHI effect will see stronger response to high temperatures
and weaker response to low temperatures.
Current studies on the mechanism and magnitude of UHIs effect on REU are
limited to theoretical discussion [Yow, 2007, Stone Jr and Rodgers, 2001] and specific
cases [Li et al., 2011, Giridharan et al., 2004]. This study adds to the literature by
62
examining the effect of UHI directly and quantitatively, exploiting granular house-
hold level data and large variations among all the major determinants of electricity
use. This paper is among the first to include fine geographical information that
reflects the urban heat island intensity that are specific to household locations.
My empirical analysis shows that urban heat island does have a significant
effect on residential electricity use, through interacting with local weather conditions
such as temperature. Higher UHI increases residential electricity consumption by
increasing the impact of Degree Days on REU. The extent of this interaction varies
seasonally and regionally. The analysis also shows dwelling attributes might change
the impact of urban head island intensity.
Focusing the empirical analysis on China has its additional significance. With
accelerating urbanization and rising income levels in China, residential energy con-
sumption has grown rapidly over the last two decades, with an annual growth rate
of 8 percent. China experienced a doubled REU per capita and a 10% increase of
urban population over the last decade, making residence the second largest sector
in total energy consumption (11%) in 2012, after industry [of Stastistics, 2014]. The
growth in demand for electricity, in particular, is even higher, with an annual aver-
age growth rate of 12.35 percent (NBS, 2012). This rapid increase mirrors Chinas
fast urbanization.
In the near future of 2030, with 60 percent of the population expected to live
in cities, residential energy consumption is likely to continue its rapid growth. The
strong energy demand, on one hand, reflects the improvement of people?s quality of
life and the nation?s development level (Niu et al., 2012). On the other hand, With
63
a coal-dominated electricity production sector, such trend will no doubt add to the
pressure on controlling the aggravating particulate and SO2 pollution, fulfilling the
commitment of CO2 abatement in Paris Accord as well as achieving the goal of
low-carbon city as a part of the blueprint of the New-type Urbanization proposed
by the Chinese government. Facing these challenges, understanding whether and
how green urban development can be a useful tool for moderating residential energy
consumption becomes crucial.
The rest of this chapter is organized as as follows. Section 2 describes the data
used in the empirical analysis. Section 3 describes the econometric approach taken.
In section 4, I discuss the results. Section 5 concludes.
3.2 Data
3.2.1 MODIS
Variables related to urban heat island intensity and local weather conditions
mainly come from MODIS dataset of NASA 2 . MODIS (or Moderate Resolution
Imaging Spectroradiometer) is a key instrument aboard the Terra (originally known
as EOS AM-1) and Aqua (originally known as EOS PM-1) satellites. Terra?s orbit
around the Earth is timed so that it passes from north to south across the equator
in the morning, while Aqua passes south to north over the equator in the afternoon.
Terra MODIS and Aqua MODIS are viewing the entire Earth?s surface every 1 to
2 days, acquiring data in 36 spectral bands between 0.405 and 14.385 m, and it
2https://modis.gsfc.nasa.gov
64
acquires data at three spatial resolutions ? 250m, 500m, and 1,000m.
The data I use includes monthly land surface temperature (from Land Surface
Temperature product MOD11A2), annual land cover type (from Land Cover Prod-
uct MCD12Q1), annual tree cover percentage (from Vegetation Continuous Fields
product MOD44B) and annual evapotranspiration (from Evapotranspiration prod-
uct MOD16A3). Land Surface Temperature product MOD11A2 has readings for
daytime and nighttime. I use the average to create HDD and CDD for each house-
hold at monthly level, which is closely related to the needs for cooling and heating.
Figure 3.1 presents the temporal variation of HDD and CDD through out the year.
1 2 3 4 5 6 7 8 9 10 11 12
CDD HDD
Figure 3.1: Box plot of CDD and HDD by month
Land Cover Product (MCD12Q1) labels the surface by land cover type. The
65
0 10 20 30 40 50
legend I use is developed by researcher at University of Maryland and has 15 cate-
gories in total, such as Forest, Grassland, Cropland, Wetland/Water and Built-up
land (see Table ). Among those, the larges group in my sample is built-up land,
with 63 percent of the total households living in an area that covered by mostly
built up land, which is typical in urban areas). In the analysis, I combine some of
the smaller groups that are similar to each other together. (see Table ).
Figure 3.2: Land Cover Type legend and class definitions
Vegetation Continuous Fields product (MOD44B) describes how much of the
66
surface within a grid is covered by trees (Tree cover percentage), non-tree vegetation
and non vegetation. All three numbers sum up to 100.
MODIS dataset divides the earth surface into grids at different resolution.
Even the most coarse grid at 1km by 1km would still provide much more refined
information about surface temperature and other land attributes related to UHI. By
matching these data with households data using household specific GPS, I create an
dataset that makes assessing the relationship between UHI and REU at the granular
level possible.
3.2.2 CRECS 2012
The household level energy consumption (electricity) and social economics
data comes from China Residential Energy Consumption Survey (CRECS) in 2012.
(cite) The CRECS survey was administrated by the Department of Energy Eco-
nomics at Renmin University of China during February 2013. It covers 26 provinces
of China. A total of 1640 households were initially invited to take the survey, while
1542 eventually enrolled in the study (a high response rate of 94 percent). After
validity and consistency checks, 1450 total observations remained for the final anal-
ysis. (see figure for location of households). Figure 3.3 shows the location variation
of households in the survey.
The survey is heavily based on the US Department of Energy and Energy Infor-
mation Administration Residential Energy Consumption Survey. It covers six areas:
household demographics, dwelling characteristics, household appliances, space heat-
67
ing and cooling, patterns of private transportation, and electricity billing, metering,
and pricing options.
Figure 3.3: Households in CRECS 2012
Figure 3.4 (Source: [Zheng et al., 2014]) provides brief profiles of the household
characteristics of the surveyed observations, in comparison with the official National
Bureau of Statistics number shown in the three right columns.
68
69
Figure 3.4: Profiles of household characteristics comparing CRECS2012 and NBS2013
In CRECS, the primary variable of interest for this study is household electric-
ity use. It is for consumption purposes only, rather than for production purposes and
it covers every month in 2012. Electricity bills or electricity consumption records
for 2012 are verified for households to be included in the survey, which ensures the
reliability of the data.
In addition to electricity use, I also include other household characteristics from
the survey as explanatory variables and controls for electricity use in the analysis.
Those include family size, income, dwelling built year, dwelling size, dwelling type,
urban/suburban/rural dummy, ownership. The summary statistics are presented in
Table 3.1 to 3.6.
70
71
Table 3.1: Summary Statistics
(1)
mean sd count p10 p25 p50 p75 p90 p99 min max
Electricity Consumption 158.1718 152.232 15104 52 80 120 193 298 685 0 9345
Household Size 2.64592 1.062632 17352 2 2 2 3 4 6 1 8
Degree Days 306.3115 230.4514 16714 51.86262 142.4704 271.6987 395.7219 627.8995 1110.99 .0303841 1561.56
NonVege Percentage 45.62288 17.5414 16992 20 32.5 50 60 65 77 3 91
Observations 17400
Table 3.2: Income
(1)
Income
10k yuan Freq. Percent CumPct
?1 492 2.85 2.85
(1-3] 2496 14.46 17.32
(3-5] 3732 21.63 38.94
(5-8] 3840 22.25 61.20
(8-12] 3108 18.01 79.21
(12-15] 1056 6.12 85.33
(15-20] 1104 6.40 91.72
(20-25] 660 3.82 95.55
(25-30] 168 0.97 96.52
(30-35] 108 0.63 97.15
(35-40] 108 0.63 97.77
(40-45] 60 0.35 98.12
(45-50] 120 0.70 98.82
(50-80] 72 0.42 99.24
(80-120] 72 0.42 99.65
(120-200] 48 0.28 99.93
(200-500] 12 0.07 100.00
Total 17256 100.00
From Figure 3.3 and Table 3.1 to 3.6 , we could see substantial variation across
household geographically and in social economics characteristics.
3.3 Empirical Strategy
The main hypothesis is that urban heat island is related to residential electric-
ity use through shifting the level of cooling and heating needed as well as through
interacting with local climatic conditions such as temperature. Higher UHI increases
electricity use through higher anthropogenic heat emission and lower evapotranspi-
ration and this effect is enhanced with higher temperature . For example, places
covered mostly by built-up surface such as buildings and asphalt generally will en-
72
Table 3.3: Dwelling Szie
(1)
Dwelling Size
Freq. Percent CumPct
? 12m2 36 0.21 0.21
(12-30] m2 60 0.35 0.55
(30-50] m2 612 3.53 4.09
(50-70] m2 1944 11.22 15.30
(70-90] m2 3132 18.07 33.38
(90-120] m2 4620 26.66 60.04
(120-150] m2 3900 22.51 82.55
(150-180] m2 888 5.12 87.67
(180-250] m2 1344 7.76 95.43
> 250m2 792 4.57 100.00
Total 17328 100.00
Table 3.4: Built Year
(1)
Built Year
Freq. Percent CumPct
Before 1949 96 0.55 0.55
1949-1959 36 0.21 0.76
1960-1969 132 0.76 1.52
1970-1979 384 2.22 3.74
1980-1989 2196 12.68 16.42
1990-1999 5148 29.73 46.15
2000-2009 8400 48.51 94.66
After 2010 924 5.34 100.00
Total 17316 100.00
joy higher UHI than rural places covered by vegetation. It also means hot days
will be ?hotter? with higher UHI. Electricity consumption will be more responsive
to weather conditions since people increase cooling which in turn increases anthro-
pogenic heat emission.
Therefore the main specification (Equation 3.13.2 )features variables that re-
flects UHI intensity such as Land Cover Type and Non Vegetation Cover Percentage
and their interactions with Degree Day, a major determinant of REU.
73
Table 3.5: Urban
(1)
Urban
Freq. Percent CumPct
City 11136 64.18 64.18
Town 2820 16.25 80.43
Rural 3396 19.57 100.00
Total 17352 100.00
Table 3.6: Ownership
(1)
Ownership
Freq. Percent CumPct
Own 16104 93.19 93.19
Rent 1176 6.81 100.00
Total 17280 100.00
Eit = ?0 + ?1DDit + ?2LCi + ?DDit ? LCi + ?Xit + ?i + ?it (3.1)
Eit = ?0 + ?1DDit + ?3NV Ci + ?DDit ?NV Ci + ?Xit + ?i + ?it (3.2)
where Eit is the electricity use for household i in month t, DDit is degree
days (insert), LCi is a categorical variable denoting land cover types, NV Ci is a
continuous variable denoting the percentage of area not covered by any vegetation
or trees, Xit represents time-varying household characteristics and ?i is household
dummy.
The identification strategy of this paper relies on the assumption that weather
conditions and land cover type could be viewed as exogenous to residential electricity
use after controlling for household characteristics. Households are generally not
sorting based on their energy consumption needs to certain types of locations and
74
even less so within a year.
If household level controls stay constant throughout the year, individual fixed
effect term would absorb their effects. For robustness checks, I also use the specifi-
cation with month and province fixed effects with household-level control variables.
Eit = ?0 + ?1DDit + ?3NV Ci + ?DDit ?NV Ci + ?Xit +mt + pi + ?it (3.3)
3.4 Results
3.4.1 Overall
Table 3.7 presents the results from running the main specification (Equation
3.1 and Equation 3.2) using a household fixed effect model. First, Degree Days is
positively correlated with electricity use with a coefficient of 0.0368 and the effect
is strongly statistically significant. This means for the baseline group (which com-
prises households living in urban/built up areas) on average a one unit increase in
monthly Degree Days would translate to about 0.04 kWh increase in the household?s
electricity consumption. The sign of this coefficient is inline with expectation.
In column one, the variables of interest are the interaction terms between Land
Cover Type and Degree Days. The coefficients represent the additional impact of
Degree Days on residential electricity use for each land cover type, in relative to the
baseline group (which is Urban/Build-up, also the largest group).
Since household level fixed effect is included, the time-invariant variables such
as Land Cover Type and other household controls such as annual income, household
75
size, etc are absorbed in flexed effect dummy and omitted from the regression. We
do not get the baseline effect of Land Cover Type, only its interaction effect with
Degree Days.
Among those, the interaction term for grassland is significant at 0.001 level
and forest and cropland is significant at 0.05 level. The coefficients for Land Cover
classified as water and barren are not statistically significantly different from zero.
The number of households in those two groups are very small, which probably
explains the non-significance of coefficients.
Both grassland and cropland have a negative coefficient for the interaction
term. This shows the magnitude of effect of Degree Days on electricity use is re-
duced when the household is surrounded mostly by grassland or cropland. This
partially confirm the hypothesis that lower urban heat island effect means house-
holds are less responsive in their electricity use to weather conditions. Specifically,
electricity use increase from one unit increase in Degree Days is reduced by ?0.0173
from 0.0368kWh to about 0.02kWh in cropland and it is reduced by ?0.0304 from
0.0368kWh to about 0.006kWh in grassland.
The positive coefficient for the interaction between forest and Degree Days is
surprising. It represent an approximately 70 percent increase in the responsiveness
to Degree Days for an average household surrounded by forest type area relative to
an average household living in built-up area. The coefficient is just significant at
0.05 level. I expected forest/scrubland/savanna area would demonstrate the least
urban heat island effect and the coefficient would be at least similar to the ones for
grassland or cropland. However the sign is flipped.
76
The positive coefficient might be due to unobserved correlation between forest
land cover and the type of households that are more responsive to Degree Days. It
could also be due to confounding variables that are driving them both in the same
direction. This is explored in subsequent analysis.
The second column corresponds to the results from Equation 3.2. Degree
Days is still positively significantly correlated with electricity consumption, with a
magnitude of 0.0194. Again, since Non-Vegetation Cover Percentage is an annual
variable, it does not vary throughout the year for households and the level effect is
absorbed by the fixed effect term. The variable of interest here is the interaction of
Degree Days and Non Vege Percentage. The coefficient is 0.000241, indicating for
an area with mean non-vegetation cover of 46 percent, the effect of Degree Days is
increased by about 0.01, about 50 percent increase relative to baseline. The positive
sign is consistent with the expectation that less vegetation cover means higher urban
heat island intensity. However, the coefficient is not statistically significant with a
p-value of 0.16.
3.4.2 Summer
Since urban heat island effect is correlated with weather conditions and it is
expected to affect heating and cooling needs differently, it is worthwhile to explore
the seasonal variation in the impacts of UHI on REU. Here, I run the analysis
separately for the summer and winter months of the year.
Table 3.8 presents the results from running the main specification (Equation
77
Table 3.7: Regression Table
(1) (2)
electricity consumption electricity consumption
Degree Days 0.0368??? 0.0194?
(9.19) (2.33)
Urban/Built up ? Degree Days 0
(.)
forest/shrubland/savanna ? Degree Days 0.0275?
(1.98)
grassland ? Degree Days ?0.0304???
(?3.42)
cropland ? Degree Days ?0.0173?
(?2.07)
water ? Degree Days 0.0435
(0.84)
barren ? Degree Days ?0.0247
(?0.80)
Degree Days ? NonVege 0.000241
(1.41)
Constant 143.9??? 145.0???
(122.95) (126.07)
Observations 14477 14375
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
78
3.1 and Equation3.2) using just observations from April to September.
First, the results in summer month share the general pattern with the ones for
the whole year. The sign of key coefficients and significance level are similar, but
the magnitude overall is much larger.
Degree Days coefficient comes in at 0.147 and 0.153 respectively. For the first
specification, it increases by a factor of four compared with results using the en-
tire year data. This indicates the fact that electricity energy consumption is more
responsive to weather conditions in the summer, probably since cooling activity is
driving the consumption. The magnitude of interaction terms also increase dramat-
ically, from ?0.0304 and ?0.0173 to 0.127 and 0.0583 for Grassland and Cropland.
This supports the hypothesis that UHI impacts electricity use in different ways in
summer and winter.
The coefficient of interaction term between Forest and Degree Days is still
positive and significant, suggesting seasonal variation may not be driving the results.
3.4.3 Winter
Table 3.9 shows the results from winter months (Equation 3.1 and Eq 3.2)
using observations from the rest of the year (October to March).
Degree Days coefficient comes in at 0.0260 in column one. For the first specifi-
cation, it much smaller than the summer and smaller than the overall results . This
indicates the fact that electricity energy consumption is much less responsive to
weather conditions in the winter. One probable explanation is that a lot households
79
Table 3.8: Summer
(1) (2)
electricity consumption electricity consumption
Degree Days 0.147??? 0.153???
(14.47) (5.75)
Urban/Built up ? Degree Days 0
(.)
forest/shrubland/savanna ? Degree Days 0.0701?
(2.33)
grassland ? Degree Days ?0.127??
(?2.88)
cropland ? Degree Days ?0.0583?
(?2.27)
water ? Degree Days 0.0887
(0.89)
barren ? Degree Days ?0.128
(?1.32)
Degree Days ? NonVege ?0.000275
(?0.52)
Constant 126.9??? 127.2???
(60.90) (60.86)
Observations 7295 7244
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
80
turn to other energy sources for heating needs.
Now coefficients of both grassland and cropland are not statistically significant
any more. The magnitude also reduces and is close to zero. This could be partially
explained by the fact most of grassland and cropland are covered by annual plants.
In the winter, the urban heat island intensity for these areas is probably not much
different from urban/built up land.
Table 3.9: Winter
(1) (2)
electricity consumption electricity consumption
Degree Days 0.0260??? 0.0193
(4.69) (1.51)
Urban/Built up ? Degree Days 0
(.)
forest/shrubland/savanna ? Degree Days 0.0428?
(2.32)
grassland ? Degree Days ?0.0108
(?0.73)
cropland ? Degree Days 0.00940
(0.79)
water ? Degree Days ?0.0229
(?0.37)
barren ? Degree Days 0.000109
(0.00)
Degree Days ? NonVege 0.000200
(0.77)
Constant 139.0??? 139.4???
(72.10) (72.21)
Observations 7182 7131
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
81
3.4.4 North vs South
To further explore the relationship between UHI and REU, I split the sample
into North China and South China and run the analysis separately for each half.
The result from examining summer time and winter time separately suggest
that most of the impact of Degree Days on REU is driven by cooling demand in
summer months, while electricity consumption is less affected by cold weather in
the winter. Since most parts of North China have central heating and the summer
is generally not as hot or humid as South China, the overall impact of Degree Days
on electricity consumption is expected to be smaller in Nouth China.
Table 3.10 and Table 3.11 present the results. The empirical numbers confirm
the argument above. The coefficient of Degree Days (0.0168) from the north is about
10 times smaller than the one from the south (0.160). Both of them are strongly
statistically significant.
If we look at the rest of the coefficients from the North regression, all the
interaction term become statistically insignificant. Furthermore, the interaction
term between Forest land and Degree Days show the expected sign.
The coefficients from the South do show Forest land and Cropland negatively
interact with Degree Days, reduce its marginal impact on REU and the effect are
highly significant.
82
Table 3.10: North
(1) (2)
electricity consumption electricity consumption
Degree Days 0.0168??? 0.00298
(4.68) (0.39)
Urban/Built up ? Degree Days 0
(.)
forest/shrubland/savanna ? Degree Days ?0.0162
(?0.80)
grassland ? Degree Days ?0.0104
(?1.38)
cropland ? Degree Days ?0.00561
(?0.75)
barren ? Degree Days ?0.00465
(?0.18)
Degree Days ? NonVege 0.000231
(1.49)
Constant 137.5??? 137.4???
(109.14) (108.99)
Observations 8077 8035
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
83
Table 3.11: South
(1) (2)
electricity consumption electricity consumption
Degree Days 0.160??? 0.0554?
(12.79) (2.40)
Urban/Built up ? Degree Days 0
(.)
forest/shrubland/savanna ? Degree Days ?0.0674??
(?3.02)
cropland ? Degree Days ?0.0918???
(?3.58)
water ? Degree Days ?0.0792
(?1.29)
Degree Days ? NonVege 0.00166??
(3.23)
Constant 139.6??? 140.4???
(57.72) (57.93)
Observations 6400 6340
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
84
3.4.5 Additional Analysis
The paper also exploits the information about dwelling structure in the survey
to examine the relationship between characteristics of the dwelling and electricity
use. The analysis explores if certain housing attributes mitigate the impact of heat
island effect to some extent. Presumably, houses that are better insulated ( newer
buildings) are less affected by the UHI. Also, detached single house and apartment
building might behave differently to UHI effect. The results are presented in Table
3.12 and 3.13.
3.5 Conclusion
This study is among the first to use micro-level data in a developing country to
study the impact of urban heat island on residential electricity use. Using household
specific location, I combine data from a household energy consumption survey in
China with detailed information related to urban heat island intensity and weather
conditions derived from remote sensing.
My empirical analysis shows that urban heat island does have a significant
effect on residential electricity use, through interacting with local weather conditions
such as temperature. Higher UHI increases residential electricity consumption by
increasing the impact of Degree Days on REU. The extent of this interaction varies
seasonally and regionally. The analysis also shows dwelling attributes might change
the impact of urban head island intensity.
UHI not only affects energy consumption, but also alters eco-environments
85
Table 3.12: Single Family Home and Uran Home
(1) (2)
electricity consumption electricity consumption
Degree Days 0.0390??? 0.0365???
(9.38) (8.52)
single family home=0 ? Degree Days 0
(.)
single family home=1 ? Degree Days ?0.0153
(?1.95)
Urban/Built up ? Degree Days 0 0
(.) (.)
forest/shrubland/savanna ? Degree Days 0.0317? 0.0311?
(2.26) (2.20)
grassland ? Degree Days ?0.0298??? ?0.0263??
(?3.34) (?2.91)
cropland ? Degree Days ?0.0117 ?0.00677
(?1.32) (?0.74)
water ? Degree Days 0.0457 0.0488
(0.88) (0.94)
barren ? Degree Days ?0.0235 ?0.0244
(?0.76) (?0.79)
city ? Degree Days 0
(.)
town ? Degree Days 0.0144
(1.35)
rural ? Degree Days ?0.0245?
(?2.47)
Constant 144.1??? 144.0???
(122.66) (122.13)
Observations 14477 14441
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
86
Table 3.13: Built after 1990
(1)
electricity consumption
Degree Days 0.0282??
(3.28)
new=0 ? Degree Days 0
(.)
new=1 ? Degree Days 0.00995
(1.14)
Urban/Built up ? Degree Days 0
(.)
forest/shrubland/savanna ? Degree Days 0.0288?
(2.07)
grassland ? Degree Days ?0.0299???
(?3.36)
cropland ? Degree Days ?0.0165?
(?1.96)
water ? Degree Days 0.0465
(0.90)
barren ? Degree Days ?0.0260
(?0.84)
Constant 143.9???
(122.63)
Observations 14422
t statistics in parentheses
? p < 0.05, ?? p < 0.01, ??? p < 0.001
87
such as biodiversity, water and air quality as well as human health and well-beings
like an increase in morbidity, mortality, and risk of violence [Zhou et al., 2014, 2015].
Combining the physical process of UHI with the socio-economic dynamics within the
urban area, this study will provide policy implications for urban population manage-
ment, building environment design, infrastructure investment, land use regulation
as well as climate change mitigation. For China and other countries experiencing
rapid urbanization, this study will inform an urban development plan with smaller
carbon footprint before the high-carbon urban infrastructure is locked in.
88
Appendix A: Chapter 1
A.1 A: Analytical Derivations
A.1.1 deriving equation 1.6
From household?s optimization conditions, we have
?u ?u ?u
= ?Px; = ?Py; = ?(1? tL) (A.1)
?X ?Y ?l
Plug them into the equation above
1 dU dX dY dl ?? dD
= Px + Py + (1? tL) ? (A.2)
? d?D d?D d?D d?D ? d?D
From differentiating final firm production function, we obtain
?X ?X
dX = dCx + dDx (A.3)
?Cx ?Dx
Also, from the first-order conditions for profit maximization, the marginal
89
products equal the input price divided by the product price, or
?X qC ?X qD + ?D
= ; = (A.4)
?Cx Px ?Dx Px
Substituting that in, multiplying through by Px gives the expression
PxdX = qCdCx + (qD + ?D)dDx (A.5)
Similarly,
PydY = qCdCy + (qD + ?D)dDy (A.6)
Adding those two equations up,
PxdX + PydY = (qD + ?D)(dDx + dDy) + qC(dCx + dCy)
= (qD + ?D)(dD) + qC(dC)
= ?DdD + qDdD + qCdC
= ?DdD + dLD + dLC
= ?DdD + dL
Substituting that in Eq(A.2 )
1 dU ? ?
? dD dL
= (?D ) + tL (A.7)
? d?D ? d?D d?D
90
A.1.2 deriving equation 1.7
Since L = L(tL, ?D),
dL ?L ?L dtL
= + (A.8)
d?D ??D ?tL d?D
Totally differentiating government budget gives
dG dD dtL dL
= D + ?D + L + tL (A.9)
d?D d?D d?D d?D
With two equation above, we could get
dL dD ?L
tL = Z(D + ?D ) + (Z + 1)tL (A.10)
d?D d?D ?tD
where
?t ?LL ?t
Z = L (A.11)
L+ t ?LL ?tL
So
1 dU ??? dD dD ?L= ?(?D ?? ) ? + Z? (D +???D? d?D ? d?D d?D?) + (?Z + 1?)?tL (A.12)?tD?
dW p ?WR ?W I
A.1.3 deriving equation 1.11
From differentiating the firm production function, we obtain
?Y ?Y
dY = dCY + dDY (A.13)
?CY ?DY
91
Also, from the first-order conditions for profit maximization, the marginal
products equal the input price divided by the Lagrangian multiplier ?, ? represents
marginal cost of producing Y
?Y qC ?Y qD + ?D
= ; = (A.14)
?CY ? ?DY ?
Plug them in the equation above, we have
?dY = qCdCy + (qD + ?D)dDy (A.15)
PydY = (Py ? ?)dY + qCdCy + (qD + ?D)dDy (A.16)
The firm producing X still holds this
PxdX = qCdCx + (qD + ?D)dDx (A.17)
So we have
PxdX + PydY = (qD + ?D)(dDx + dDy) + qC(dCx + dCy) + (Py ? ?)dY
= ?DdD + dL+ (Py ? ?)dY
Consider the welfare effect of a policy change in ?D, we still have
1 dU dX dY dl ??? ? dD= Px + Py + (1 tL) (A.18)
? d?D d?D d?D d?D ? d?D
92
Plug PxdX + PydY in there
1 dU ? ?
? dD dY dL
= (?D ) + (Py ? ?) + tL (A.19)
? d?D ? d?D d?D d?D
A.1.4 deriving equation 1.13
Since L = L(tL, ?D, Py)
dL ?L ?L dtL ?L dPy
= + + (A.20)
d?D ??D ?tL d?D ?Py d?D
Government budget is G = I + tLL+ ?DD Differentiate
dG dI dD dtL dL
= +D + ?D + L + tL (A.21)
d?D d?D d?D d?D d?D
Combining the two equations, we could get
dL dI dD ?L ?L dPy
tL = Z( +D + ?D ) + (Z + 1)tL + (Z + 1)tL (A.22)
d?D d?D d?D ?tD ?Py d?D
When Py = P?y, dPy = 0 So we have
1 dU ? ?
? dD ? dY dI dD ?L ?L dPy= (?D ) + (Py ?) + Z( +D + ?D ) + (Z + 1)tL + (Z + 1)tL
? d?D ? d?D d?D d?D d?D ?tD ?Py d?D
(A.23)
? ?
? dD ? dY dI dD ?L= (?D ) + (Py ?) + Z( +D + ?D ) + (Z + 1)tL
? d?D d?D d?D d?D ?tD
(A.24)
93
A.2 B: the Numerical Model
A.2.1 Parameters
Firm Behavior Parameters
aj productivity parameter for production of good j
?ij distribution parameter for input i in production of good j
?j substitution parameter for production of good j (note: ? = (??1)/? and
? is the elasticity of substitution between factors in production.)
Household Behavior Parameters
L? total labor endowment
?i, ?l distribution parameter for utility function
?u substitution parameter for utility function
Government Policy Parameters
Gh government transfers to households in real terms
Gi government spendings on good i in real terms
Emission Parameters
bj emissions of carbon dioxide per unit of good j produced
A.2.2 Equations
?
? ?
Xj = aj( ?ijX
j + ? j 1/?jij LjLj ) (A.25)
i
94
?
U = ( ? C?u + ? ?u 1/?ui i ll ) ? ?(e) (A.26)
i
The government?s budget constraint is given by:
? ?
piGi + pcGH = ?LpLL+ ?c piCi + ?te (A.27)
i i
With price regulation for electricity sector, the government?s budget constraint is :
? ? ?
piGi + pcGH = ?LpLL+ ?c piCi + ?te+ ?i (A.28)
i i
95
Bibliography
Sohail Ahmad, Giovanni Baiocchi, and Felix Creutzig. Co2 emissions from direct
energy use of urban households in india. Environmental Science & Technology,
49(19):11312?11320, 2015.
Maximilian Auffhammer and Erin T. Mansur. Measuring climatic impacts on energy
consumption: A review of the empirical literature. Energy Economics, 46:522?
530, November 2014. ISSN 0140-9883. doi: 10.1016/j.eneco.2014.04.017. URL
http://www.sciencedirect.com/science/article/pii/S0140988314001017.
R. W. Bacon and J. Besant-Jones. Global electric power reform, pri-
vatization, and liberalization of the electric power industry in develop-
ing countries. Annual Review of Energy and the Environment, 26(1):331?
359, November 2001. ISSN 1056-3466. doi: 10.1146/annurev.energy.26.1.
331. URL https://www-annualreviews-org.proxy-um.researchport.umd.
edu/doi/10.1146/annurev.energy.26.1.331.
Antonio M. Bento and Mark Jacobsen. Ricardian rents, environmental policy
and the double-dividend hypothesis. Journal of Environmental Economics and
Management, 53(1):17?31, January 2007. ISSN 0095-0696. doi: 10.1016/
j.jeem.2006.03.006. URL http://www.sciencedirect.com/science/article/
pii/S0095069606000775.
John E Besant-Jones. Reforming power markets in developing countries: what have
we learned? The World Bank ; Energy and Mining Sector Board, Washington,
DC; Washington, DC, 2006. OCLC: 211773955.
A. L. Bovenberg and F. van der Ploeg. Environmental policy, public finance and the
labour market in a second-best world. Journal of Public Economics, 55(3):349?
390, November 1994. ISSN 0047-2727. doi: 10.1016/0047-2727(93)01398-T. URL
http://www.sciencedirect.com/science/article/pii/004727279301398T.
A. Lans Bovenberg and Lawrence H. Goulder. Optimal Environmental Taxation
in the Presence of Other Taxes: General- Equilibrium Analyses. The American
Economic Review, 86(4):985?1000, September 1996. URL http://www.jstor.
org/stable/2118315.
96
A. Lans Bovenberg and Lawrence H. Goulder. Chapter 23 Environmental taxa-
tion and regulation. In Alan J. Auerbach and Martin Feldstein, editor, Hand-
book of Public Economics, volume Volume 3, pages 1471?1545. Elsevier, 2002.
ISBN 1573-4420. URL http://www.sciencedirect.com/science/article/
pii/S1573442002800271.
A. Lans Bovenberg and Ruud A. de Mooij. Environmental Levies and Distortionary
Taxation. The American Economic Review, 84(4):1085?1089, September 1994.
URL http://www.jstor.org/stable/2118046.
Jared C. Carbone and V. Kerry Smith. Evaluating policy interventions with general
equilibrium externalities. Journal of Public Economics, 92(56):1254?1274, June
2008. ISSN 0047-2727. doi: 10.1016/j.jpubeco.2007.06.003. URL http://www.
sciencedirect.com/science/article/pii/S0047272707000849.
Helmuth Cremer, Firouz Gahvari, and Norbert Ladoux. Environmental tax de-
sign with endogenous earning abilities (with applications to France). Journal of
Environmental Economics and Management, 59(1):82?93, January 2010. ISSN
0095-0696. doi: 10.1016/j.jeem.2009.04.009. URL http://www.sciencedirect.
com/science/article/pii/S0095069609000850.
Lucas W Davis and Paul J Gertler. Contribution of air conditioning adoption to
future energy use under global warming. Proceedings of the National Academy of
Sciences, 112(19):5962?5967, 2015.
Reid Ewing and Fang Rong. The impact of urban form on us residential energy use.
Housing Policy Debate, 19(1):1?30, 2008.
Carla Freeman and Bo Kong. Making Sense of Carbon Market Development.pdf.
Carbon & Climate Law Review, 7(3):194, September 2013.
Don Fullerton and Seung-Rae Kim. Environmental investment and policy with
distortionary taxes, and endogenous growth. Journal of Environmental Economics
and Management, 56(2):141?154, September 2008. ISSN 0095-0696. doi: 10.1016/
j.jeem.2008.02.001. URL http://www.sciencedirect.com/science/article/
pii/S0095069608000508.
Renganathan Giridharan, Sivaguru Ganesan, and SSY Lau. Daytime urban heat
island effect in high-rise and high-density residential developments in hong kong.
Energy and Buildings, 36(6):525?534, 2004.
Lawrence H. Goulder. Effects of Carbon Taxes in an Economy with Prior Tax Distor-
tions: An Intertemporal General Equilibrium Analysis. Journal of Environmental
Economics and Management, 29(3):271?297, November 1995. ISSN 0095-0696.
doi: 10.1006/jeem.1995.1047. URL http://www.sciencedirect.com/science/
article/pii/S0095069685710479.
97
Lawrence H. Goulder and Richard D. Morgenstern. China s Rate-based Approach
to Reducing CO 2 Emissions : Strengths , Limitations , and Alternatives. 2017.
Lawrence H. Goulder, Ian W. H. Parry, Roberton C. Williams III, and Dallas
Burtraw. The cost-effectiveness of alternative instruments for environmental
protection in a second-best setting. Journal of Public Economics, 72(3):329?
360, June 1999. ISSN 0047-2727. doi: 10.1016/S0047-2727(98)00109-1. URL
http://www.sciencedirect.com/science/article/pii/S0047272798001091.
Lawrence H. Goulder, Marc A. C. Hafstead, and Roberton C. Williams, III. General
Equilibrium Impacts of a Federal Clean Energy Standard. American Economic
Journal: Economic Policy, 8(2):186?218, May 2016. ISSN 19457731.
Lawrence H. Goulder, Richard D. Morgenstern, Clayton Munnings, and Jeremy
Schreifels. Chinas National Carbon Dioxide Emission Trading System: An Intro-
duction. Economics of Energy & Environmental Policy, 6(2), April 2017. ISSN
21605882. doi: 10.5547/2160-5890.6.2.lgou. URL http://www.iaee.org/en/
publications/eeeparticle.aspx?id=175.
Gerhard K Heilig. World urbanization prospects: the 2011 revision. United Na-
tions, Department of Economic and Social Affairs (DESA), Population Division,
Population Estimates and Projections Section, New York, 2012.
Mun Ho, Zichao Yu, and Zhongmin Wang. Chinas Power Generation Dis-
patch, April 2017. URL https://www.rff.org/publications/reports/
chinas-power-generation-dispatch/.
David Pollitt Michael Tooraj Mota Jamasb, Raffaella Newbery. Electricity Sector
Reform in Developing Countries : A Survey of Empirical Evidence on Deter-
minants and Performance. Policy Research Working Papers. The World Bank,
March 2005. doi: 10.1596/1813-9450-3549. URL https://elibrary.worldbank.
org/doi/abs/10.1596/1813-9450-3549.
Christopher Jones and Daniel M Kammen. Spatial distribution of us household
carbon footprints reveals suburbanization undermines greenhouse gas benefits of
urban population density. Environmental science & technology, 48(2):895?902,
2014.
Madhu Khanna and Narasimha D. Rao. Supply and Demand of Electricity
in the Developing World. Annual Review of Resource Economics, 1(1):567?
596, September 2009. ISSN 1941-1340. doi: 10.1146/annurev.resource.050708.
144230. URL https://www-annualreviews-org.proxy-um.researchport.
umd.edu/doi/10.1146/annurev.resource.050708.144230.
Sungwon Lee and Bumsoo Lee. The influence of urban form on ghg emissions in
the us household sector. Energy Policy, 68:534?549, 2014.
98
Junxiang Li, Conghe Song, Lu Cao, Feige Zhu, Xianlei Meng, and Jianguo Wu.
Impacts of landscape structure on surface urban heat islands: a case study of
shanghai, china. Remote Sensing of Environment, 115(12):3249?3263, 2011.
Ming-Hua Liu, Dimitris Margaritis, and Yang Zhang. Market-driven coal prices
and state-administered electricity prices in China. Energy Economics, 40:167?
175, November 2013. ISSN 0140-9883. doi: 10.1016/j.eneco.2013.05.021. URL
http://www.sciencedirect.com/science/article/pii/S0140988313001096.
Kabir Malik, Maureen Cropper, Alexander Limonov, and Anoop Singh. The
Impact of Electricity Sector Restructuring on Coal-fired Power Plants in In-
dia. The Energy Journal, 36(4):287?312, 2015. ISSN 0195-6574. URL https:
//www.jstor.org/stable/24696058.
William D. Nordhaus. Revisiting the social cost of carbon. Proceedings of the
National Academy of Sciences, 114(7):1518?1523, February 2017. ISSN 0027-
8424, 1091-6490. doi: 10.1073/pnas.1609244114. URL https://www.pnas.org/
content/114/7/1518.
Chinese National Bureau of Stastistics. China statistical yearbook 2014. 2014.
Tim R Oke. The energetic basis of the urban heat island. Quarterly Journal of the
Royal Meteorological Society, 108(455):1?24, 1982.
Ian W.H. Parry, Roberton C. Williams, and Lawrence H. Goulder. When Can
Carbon Abatement Policies Increase Welfare? The Fundamental Role of Distorted
Factor Markets. Journal of Environmental Economics and Management, 37(1):
52?84, January 1999. ISSN 00950696. doi: 10.1006/jeem.1998.1058. URL http:
//linkinghub.elsevier.com/retrieve/pii/S0095069698910580.
Carlos Pombo and Rodrigo Taborda. Performance and efficiency in Colombia?s
power distribution system: Effects of the 1994 reform. Energy Economics, 28(3):
339?369, May 2006. ISSN 0140-9883. doi: 10.1016/j.eneco.2005.08.001. URL
http://www.sciencedirect.com/science/article/pii/S0140988305000794.
KC Seto and S Dhakal. Chapter 12: Human settlements, infrastructure, and spatial
planning. Climate Change 2014: Mitigation of Climate Change. Contribution of
Working Group III to the Fifth Assessment Report of the Intergovernmental Panel
on Climate Change, pages 67?76, 2014.
Brian Stone Jr and Michael O Rodgers. Urban form and thermal efficiency: how the
design of cities influences the urban heat island effect. Journal of the American
Planning Association, 67(2):186?198, 2001.
Fei Teng, Frank Jotzo, and Xin Wang. Interactions between Market Reform and
a Carbon Price in Chinas Power Sector. Economics of Energy & Environmental
Policy, 6(2), April 2017. ISSN 21605882. doi: 10.5547/2160-5890.6.1.ften. URL
http://www.iaee.org/en/publications/eeeparticle.aspx?id=177.
99
Roberton C. Williams III. Revisiting the cost of protectionism:: The role of tax
distortions in the labor market. Journal of International Economics, 47(2):429?
447, April 1999. ISSN 0022-1996. doi: 10.1016/S0022-1996(98)00023-3. URL
http://www.sciencedirect.com/science/article/pii/S0022199698000233.
Roberton C. Williams III. Environmental Tax Interactions when Pollution Af-
fects Health or Productivity. Journal of Environmental Economics and Man-
agement, 44(2):261?270, September 2002. ISSN 0095-0696. doi: 10.1006/
jeem.2001.1237. URL http://www.sciencedirect.com/science/article/pii/
S0095069601912379.
Catherine Wolfram, Orie Shelef, and Paul Gertler. How Will Energy Demand
Develop in the Developing World? Journal of Economic Perspectives, 26(1):
119?138, February 2012. ISSN 0895-3309. doi: 10.1257/jep.26.1.119. URL
https://www.aeaweb.org/articles?id=10.1257/jep.26.1.119.
Interagency Working Group on Social Cost of Carbon. Technical Support Document:
Technical Update of the Social Cost of Carbon for Regulatory Impact Analysis
under Executive Order 12866. In Social Cost of Carbon Estimates for Regulatory
Impact Analysis: Development and Technical Assessment, pages 65?88. 2016.
Lisha Yang and Boqiang Lin. Carbon dioxide-emission in Chinas power industry:
Evidence and policy implications. Renewable and Sustainable Energy Reviews,
60:258?267, July 2016. ISSN 1364-0321. doi: 10.1016/j.rser.2016.01.058. URL
http://www.sciencedirect.com/science/article/pii/S1364032116000885.
Donald M Yow. Urban heat islands: Observations, impacts, and adaptation. Geog-
raphy Compass, 1(6):1227?1251, 2007.
Junjie Zhang, Zhenxuan Wang, and Xinming Du. Lessons learned from Chinas
regional carbon market pilots. Economics of Energy & Environmental Policy,
Volume 6(Number 2), 2017a. URL https://ideas.repec.org/a/aen/eeepjl/
eeep6-2-zhang.html.
Xiaohan Zhang, Niven Winchester, and Xiliang Zhang. The future of coal in China.
Energy Policy, 110:644?652, November 2017b. ISSN 0301-4215. doi: 10.1016/
j.enpol.2017.07.001. URL http://www.sciencedirect.com/science/article/
pii/S0301421517304342.
Yin-Fang Zhang, David Parker, and Colin Kirkpatrick. Electricity sector
reform in developing countries: an econometric assessment of the effects
of privatization, competition and regulation. Journal of Regulatory Eco-
nomics, 33(2):159?178, April 2008. ISSN 0922-680X, 1573-0468. doi:
10.1007/s11149-007-9039-7. URL https://link-springer-com.proxy-um.
researchport.umd.edu/article/10.1007/s11149-007-9039-7.
100
Xinye Zheng, Chu Wei, Ping Qin, Jin Guo, Yihua Yu, Feng Song, and Zhanming
Chen. Characteristics of residential energy consumption in china: Findings from
a household survey. Energy Policy, 75:126?135, 2014.
Decheng Zhou, Shuqing Zhao, Shuguang Liu, Liangxia Zhang, and Chao Zhu. Sur-
face urban heat island in china?s 32 major cities: Spatial patterns and drivers.
Remote Sensing of Environment, 152:51?61, 2014.
Decheng Zhou, Shuqing Zhao, Liangxia Zhang, Ge Sun, and Yongqiang Liu. The
footprint of urban heat island effect in china. Scientific reports, 5, 2015.
101