Categorization of a Residential Area FSR (BC)

Assessment of Residential
Electricity Demand in Korea
using Aggregated Panel Data
Jiyeon Kim
Korea University
1
Contents
01 Introduction
02 Model
03 Data and Variables
04 Estimation Results
05 Forecasting Accuracy
06 Concluding Remarks
2
01
Introduction
3
Introduction
•
Recently, the problem of the electricity shortage has emerged as a serious problem in Korea.
•
The blackout in September 15, 2011 is the biggest example that indicates the insecurity of the
electricity supply system in Korea.
•
Still later, the electricity reserves in Korea plummet to an alarming level whenever the demand
of electricity unexpectedly jumps due to the intense heat or cold.
Trend of electricity backup power and
reserve rate
Trend of electricity consumption
500,000,000
900
800
700
600
500
400
300
200
100
0
450,000,000
400,000,000
350,000,000
300,000,000
250,000,000
200,000,000
Electricity Consumption (MWH)
4
18
16
14
12
10
8
6
4
2
0
backup power (10,000kW)
electric power reserve rate (%)
Introduction
•
In order to solve this problem by implementing appropriate policies, it is far more important to
figure out determinants of the electricity demand and calculate the magnitude of their impact, in
other words, to estimate the electricity demand equation.
•
Despite there exist extensive studies examining the residential demand for electricity, there are
few studies investigating the demand for residential electricity in Korea. Moreover, the
estimated price and income elasticities in Korea are heterogeneous.
•
Therefore, it is meaningful to provide new empirical evidence on the values of price and income
elasticities based on different methodology and data with those used in previous studies.
•
In addition, this study focuses on the impact of the socio-demographic factors such as
population aging and building structure on the electricity consumption, which have rarely been
considered in previous literatures.
5
Introduction
Literatures on the relation between population aging and the energy consumption
•
Dalton et al. (2008) : suggested that population aging reduced energy consumptions and
emissions in the long run.
•
Kronenberg (2009) : investigated the impact of demographic change on energy use and
greenhouse gas focusing on the fact that old people displayed different consumption patterns
with young people. He suggested that the demographic change did not reduce energy use and
greenhouse gas emissions until 2030.
•
Kim (2012) : found the inverse U-shape relationship between population aging and per capita
energy demand using panel data of 51 countries for the period of 1976-2009.
•
Won (2012) : showed population aging decreased the electricity consumption using Korean
data.
6
Introduction
Building Structure and the Energy Consumption
•
Together with economic and social factors, the structure of building also determine the energy
consumption by affecting heating and cooling capacity of a house .
•
In particular, as high-rise buildings have increased in the urban region, the interest of energy
efficiency of the tall buildings has been emerging in Korea. Nevertheless, little interest has been
paid on this issue so far.
Trend of houses
16,000,000
70
14,000,000
60
Trend of high-rise buildings
95,000
90,000
12,000,000
50
10,000,000
6,000,000
80,000
30
75,000
20
4,000,000
85,000
40
8,000,000
70,000
65,000
10
2,000,000
0
0
1990
number of houses
7
100,000
1995
2000
single-house ratio
2005
2010
apartment ratio
60,000
2005 2006 2007 2008 2009 2010 2011 2012
number of buildings above 10th floor
number of buildings above 30th floor
1,100
1,000
900
800
700
600
500
400
300
200
100
Introduction
Literatures on the relation between housing type and the electricity consumption
•
Park (1998) and Um (2006)
-
suggested that energy consumption rose as the ratio of the floor area to the surface of a building
increased.
•
Kim and Lee (2007)
-
pointed out that energy consumption represented the U-shape as the total floor area of a building
increased.
•
Choi et al. (2007)
-
analyzed the energy consumption patterns of buildings with the same volume and the various
shapes. As a result, it appeared that a single-story building as well as a skyscraper was the most
inefficient in terms of the energy consumption.
However, few studies empirically analyze the impact of house structure on electricity consumption,
especially in economics field.
8
Introduction
Most efficient
Most inefficient
Model
Energy
Consumptions for
Heating (GJ)
Energy
Consumptions for
Cooling (GJ)
Total Energy
Consumption
(GJ)
M-9
75.48
80.84
156.32
M-10
107.47
83.73
191.2
M-3
105.62
92.24
197.86
M-6
114.03
101.91
215.94
M-13
121.47
95.09
216.56
M-12
124.25
103.97
228.22
M-11
130.25
122.05
252.3
M-4
168.27
113.95
282.22
M-8
170.83
116.86
287.69
M-1
168.15
125.52
293.67
M-7
172.73
124.82
297.55
M-2
171.4
130.52
301.92
M-14
184.15
134.07
318.22
M-5
177.86
140.57
318.43
M-15
208.94
172.53
381.47
M-16
215.22
199.45
414.67
Choi et al. (2007)
9
Introduction
Purpose of this study
1. To provide new empirical evidence of price and income elasticities in Korea by
estimating the residential electricity demand equation.
2. To investigate the impact of the socio-demographic factors on the residential electricity
consumption.
10
02
Model
11
Partial Adjustment Model
Long-run equilibrium electricity demand function
ln Eit*    1 ln Pit   2 ln Yit   3 ln POPit   4 HDDit   5CDDit
  6 AGINGit   7 BC it   8 FSRit  vt  ui   it
(1)
Long-run effect
ln Eit  ln Eit 1   (ln Eit*  ln Eit 1 )   t , 0    1
(2
)
speed at which
the actual demand
converges into the equilibrium
Combining equation (1) and (2)
Final equation
Short-run effect
ln Eit    (1   ) ln Eit 1   1 ln Pit   2 ln Yit   3 ln POPit   4 HDDit   5CDDit
  6 AGINGit   7 BC it   8 FSRit  vt  ui   it
Where
12
   and  k   k
(3)
03
Data and Variables
13
Data and Variables
• This study uses the data set composed of 138 of 231 administrative districts for the
period from 2004 to 2012.
1. Electricity Variables
 E (kWh) : kWh consumed by residential consumers
(KEPCO)
 P (Won/kWh) : (KEPCO's revenues from electricity consumed in residential sector / Electricity
consumption in residential sector) / CPI * 100
(KEPCO)
 Two part tariff and the choice of the electricity price
•
•
Nordin (1976) : suggested using the marginal price as the price variable and subtracting the
fixed fee from the income.
Shin (1985) : argued that people responds to the average price.
→ Possibility of endogeneity problem
14
Data and Variables
2. Socioeconomic Variables
 Y (1,000Won) : Real GRDP in each districts (base : 2005)
 POP : Total population in each districts
(KOSIS)
(KOSIS)
 AGING (%) : Population over 65 / Total population
(KOSIS)
3. Climatic Variables
 HDD : Heating degree days with threshold of 18°C
 CDD : Cooling degree days with threshold of 24°C
(Korea Meteorological Administration)
15
Data and Variables
4. Housing Type Variables
 BC (%) : Weighted average of the standard building coverage rates of each sub residential area
(Weight : ratio of each sub residential area to the total residential area)
 FSR (%) : Weighted average of the standard floor area ratio of each sub residential area
(Weight : ratio of each sub residential area to the total residential area)
Building coverage rates : A / Lot area
C
Floor area ratio : A+B+C / Lot area
B
A
Lot area
16
(KOSIS)
Data and Variables
Categorization of a Residential Area
Residential Area
Description
General Residential Area Type 1
Low-rise housing area
General Residential Area Type 2
Middle-rise housing area
General Residential Area Type 3
Middle and high-rise housing area
Exclusive Residential Area Type 1
Single housing area
Exclusive Residential Area Type 2
Apartment housing area
Semi-residential Area
Residential and commercial area
FSR (BC) : standard FSR(BC) in General Residential Area Type 1 * ( General
Residential Area Type 1 / Total Residential Area) + standard FSR(BC) in General
Residential Area Type 2 * (General Residential Area Type 2 / Total Residential Area) +
⋯ + standard FSR(BC) in Semi-residential Area * ( Semi-residential Area / Total
Residential Area )
17
Data and Variables
Descriptive Statistics
E (kWh)
P (Won/kWh)
Y (1,000Won)
18
Mean
S.D.
Min
Max
277,000,000
267,000,000
15,000,000
1,460,000,000
120.72
10.38
98.35
170.60
5,240,000,000
6,480,000,000
217,000,000
56,200,000,000
POP
234055
216077
18208
1120258
HDD
2819.8
457.0
1026.9
6696.0
CDD
126.0
57.4
0.0
322.9
AGING (%)
14.46
6.97
4.72
33.79
BC (%)
58.68
1.84
33.78
66.15
FSR (%)
229.08
25.81
142.77
334.48
04
Estimation Results
19
Methodology
•
The inclusion of a lagged dependent variable violates the exogeneity assumption.
•
In that case, the coefficient of a lagged dependent variable becomes biased making the long-run
elasticities of electricity unreliable.
 GMM-AB (Arellano and Bond, 1991)
: takes first differences to eliminate fixed effects and use lagged instruments to correct for
simultaneity in the first-differenced equations.
 GMM-BB (Blundell and Bond, 1998)
: uses lagged first-differences as instruments for equations in level, in addition to the usual lagged
levels as instruments for equations in first-differences.
- The instruments used in the GMM-AB estimator become less informative when the variables are
close to be random-walk variables (Blundell and Bond, 1998, 2000).
- GMM-BB can consider the presence of more than one endogenous variable. (L. Blazquez et al.
2013)
- The GMM-BB approach generally produces more efficient and precise estimates compared to
GMM-AB by implementing and reducing finite sample bias (Baltagi, 2008).
20
Cons.
LnE(t-1)
LnP
LnY
LnPOP
Model1
1.2035
(0.1393)
0.7326***
(0.0146)
0.0226
(0.0252)
0.0236***
(0.003)
0.2612***
(0.0163)
-0.0247***
(0.0046)
-0.0335***
(0.004)
-0.0292***
(0.0031)
-0.0147***
(0.0026)
-0.0114***
(0.0018)
0.0226***
(0.0017)
-0.0156***
(0.0011)
0.016
-0.0223***
(0.005)
-0.0313***
(0.0045)
-0.0276***
(0.0038)
-0.0136***
(0.003)
-0.0072***
(0.0026)
0.0209***
(0.0017)
-0.0136***
(0.0013)
0.017
-0.0193***
(0.0057)
-0.0285***
(0.0047)
-0.0252***
(0.0035)
-0.0118***
(0.0029)
-0.0098***
(0.0022)
0.0244***
(0.0018)
-0.0151***
(0.0012)
0.017
-0.0218***
(0.0029)
-0.0306***
(0.0025)
-0.0276***
(0.0019)
-0.0138***
(0.0015)
-0.0112***
(0.001)
0.0224***
(0.0011)
-0.0165***
(0.0008)
0.132
Model5
1.8982
(0.1211)
0.7079***
(0.0126)
-0.0402**
(0.0204)
0.0307***
(0.0012)
0.2792***
(0.0141)
1.63E-05*
(8.82E-06)
0.0001***
(4.45E-05)
-0.0044***
(0.0006)
0.0001***
(1.8E-05)
-0.0012***
(0.0001)
-0.0023***
(0.0003)
5.48E-06***
(5.87E-07)
-1.02E-07**
(3.81E-08)
-2.55E-07
(1.86E-07)
-0.0234***
(0.0043)
-0.0322***
(0.0039)
-0.0297***
(0.0033)
-0.0156***
(0.0026)
-0.0089***
(0.0022)
0.0201***
(0.0015)
-0.0141***
(0.001)
0.254
0.003
0.003
0.003
0.002
0.002
0.389
0.375
0.389
0.329
0.312
1104
1104
1104
1104
1104
81
83
83
112
118
HDD
CDD
Model2
1.291
(0.1533)
0.7263***
(0.0154)
-0.0023
(0.0285)
0.028***
(0.0032)
0.263***
(0.0173)
-2.55E-06
(2.79E-06)
0.0001***
(1.54E-05)
Model3
1.252
(0.1948)
0.7347***
(0.0151)
-0.0047
(0.0275)
0.0273***
(0.0031)
0.2565***
(0.0173)
Model4
1.6424
(0.0485)
0.7247***
(0.0086)
-0.0092
(0.0141)
0.0275***
(0.0007)
0.2671***
(0.0086)
-0.0024**
(0.0011)
0.0001***
(2.41E-05)
AGING
AGING Square
-0.0009***
(0.0001)
-0.0022***
(0.0002)
4.52E-06***
(5.04E-07)
BC
FSR
FSR Square
HDD*FSR
CDD*FSR
_Iyear_2005
_Iyear_2006
_Iyear_2007
_Iyear_2008
_Iyear_2009
_Iyear_2010
_Iyear_2011
Hansen test (p-value)
Arellano-Bond AR1 test
(p-value)
Arellano-Bond AR2 test
21
(p-value)
Number of Observation
Number of Instruments
Estimation Results
1. The effect of price and income
Price Elasticity
Income Elasticity
Short Run
-0.04
0.03
Long Run
-0.14
0.1
•
The demand of electricity is both price-inelastic and income-inelastic.
•
Both price and income elasticities are smaller than those of previous studies.
•
The long-run elasticities are greater than the short-run elasticities.
2. The effect of population aging
•
Population aging reduces the demand of electricity, and the decreasing rate of electricity
consumption is diminishing.
•
A 1% increase of population aging drags down electricity demand by about 0.4%.
22
Estimation Results
3. The effect of climate
•
The demand of electricity increases by about 1% as CDD increases by 100 days in a year. In
contrast, it increases by about only 0.17%, if HDD increases by the same amount.
•
This may be because, in the case of heating, people can substitute electricity with the other
energy sources such as gas or kerosene oil while they usually depend on electronic appliances
such as fans or air conditioners for cooling.
Energy sources for heating
8%
Gas
Electricity
23%
49%
Oil
Others
20%
23
Estimation Results
4. The effect of building structure variables
•
The demand of electricity decreases first and then begins to increase as a building gets higher.
The calculated turning point of the floor area ratio is about 238%.
•
This empirical result supports the results of the previous studies in the field of architecture (Kim
and Lee, 2007; Choi et al., 2007) that the medium-rise house is the most energy-efficient.
Electricity Consumption
Floor Area Ratio
238%
24
Forecasting Accuracy
(lne)
2626
2624
2622
2620
2618
2616
2614
2612
2009
2010
actual
25
2011
m1
m5
2012
05
Conclusion
26
Conclusion
• This paper estimates the residential electricity demand equation on Korea based on
the panel data of 138 Korean administrative districts for the period from 2004 to 2012.
• For this purpose, we use the partial adjustment model, and GMM-BB is employed for
the estimation.
•
We find that Korean residential electricity demand is inelastic to both of the price
and income, HDD and CDD has a positive effect on the electricity consumption, and
the Population aging has a negative effect on the electricity consumption.
• Moreover, this paper investigates the influence of house structure on the electricity
demand. The results show that there exists the U-shape relation between the height of
the buildings and the electricity consumption.
27
Thank you
28