Electricity consumption and economic growth in …ageconsearch.umn.edu/bitstream/235796/2/Xenophon ppt upload.pdf · Electricity consumption and economic growth in China: assessing

Electricity consumption and economic growth in China:

assessing Granger causality at provincial, electricity-

market, and national levels

Aleksis Xenophon

The University of Adelaide

Contributed presentation at the 60th AARES Annual Conference,

Canberra, ACT, 2-5 February 2016

Copyright 2016 by Author(s). All rights reserved. Readers may make verbatim copies of this document for

non-commercial purposes by any means, provided that this copyright notice appears on all such copies.

Electricity consumption and economic growth inChina: assessing Granger causality at provincial,electricity-market, and national levels

Aleksis Xenophon

The University of Adelaide

Motivation Data & Methodology Results Conclusions

Which comes first?

1985 1990 1995 2000 2005 20100

10

20China

Agg

rega

te R

GD

PG

row

th R

ate

(%)

Year

1985 1990 1995 2000 2005 20100

10

20

Ele

ctric

ity C

onsu

mpt

ion

Gro

wth

Rat

e (%

)

RGDPELC

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Why?

I Role of electricity consumption in economic development

I Planning infrastructure

I Nationwide carbon trading scheme - 2017

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Most studies look at China as a whole

Jilin

Heilongjiang

Liaoning

Qinghai

Gansu

Shaanxi

Shanxi

Hebei

SichuanHubei

Henan

Shandong

Anhui

Jiangsu

Yunnan

Guizhou Hunan Jiangxi

Zhejiang

Hainan

Guangdong

Fujian

Guangxi

Tibet (Xizang)

Xinjiang

Inner Mongolia (Nei Mongol)

Ningxia

Chongqing

Shanghai

Tianjin

Beijing

Guangzhou

Fuzhou

Hangzhou

Nanjing

Jinan

Shijiazhuang

Shenyang

Changchun

Harbin

Xining

Lanzhou

Xi'an

Taiyuan

ChengduWuhan

Zhengzhou

Hefei

Kunming

Guiyang

ChangshaNanchang

Haikou

Ürümqi

Lhasa

Yinchuan

Hohhot

Nanning

Hong Kong SARMacau SAR

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Electricity markets in China

Tibet

Northwest

Northeast

North

EastCentral

Southern

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

No consensus in the literature

Study Direction of Granger causality

Shiu & Lam (2004) ELC⇒RGDPBoqiang (2003) RGDP⇒ELCChen et al. (2007) RGDP<ELC

Table 1 : Previous studies

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Data

I Chinese Statistical Yearbooks

I Electricity consumption and real gross regional product for 28provinces

I Excluded: Hainan and Tibet

I Aggregated: Chongqing and Sichuan

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Vector error correction model (VECM)

ELC = Log Electricity ConsumptionRGRP = Log Real Gross Regional Product

RGRPt = β0 + β1ELCt + µt

∆ELCt = α1 +αxµt−1 +∑i=1

α11(i)∆RGRPt−i +∑i=1

α12(i)∆ELCt−i + εxt

∆RGRPt = α2+αyµt−1+∑i=1

α21(i)∆RGRPt−i +∑i=1

α22(i)∆ELCt−i +εyt

Shiu and Lam (2004)

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Long run Granger causality map

ELC⇒RGRPRGRP⇒ELCRGRP⇔ELCRGRP<ELC

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Aggregation and robustness

Long run Granger causal direction[1] [2] [3] [4] [5] [6] [7] [8]

China

Southern ChinaPower Grid

GuangdongGuangxiGuizhouYunnan

ELC⇒RGRP RGRP⇒ELC RGRP⇔ELC RGRP<ELC

Model number and sample period key

[1] [2] [3] [4] [5] [6] [7] [8]

1985-2012 1985-06 1986-07 1987-08 1988-09 1989-10 1990-11 1991-12

Aleksis Xenophon The University of Adelaide

Motivation Data & Methodology Results Conclusions

Conclusions

I Regional clustering

I Limitations of highly aggregated data

I Case study for further analysis

Aleksis Xenophon The University of Adelaide

Questions

Aleksis Xenophon The University of Adelaide

References

References

Boqiang, L. (2003), ‘Structural changes, efficiency improvementand electricity demand forecasting’, Economic Research Journal5, 57–65.

Chen, S.-T., Kuo, H.-I. and Chen, C.-C. (2007), ‘The relationshipbetween GDP and electricity consumption in 10 Asiancountries’, Energy Policy 35(4), 2611–2621.

Shiu, A. and Lam, P.-L. (2004), ‘Electricity consumption andeconomic growth in China’, Energy Policy 32(1), 47–54.

Aleksis Xenophon The University of Adelaide

Co-movement

1985 1990 1995 2000 2005 20100

5

10

15

20Guangxi

Agg

rega

te R

GR

PG

row

th R

ate

(%)

Year

1985 1990 1995 2000 2005 20100

5

10

15

20

Ele

ctric

ity C

onsu

mpt

ion

Gro

wth

Rat

e (%

)

RGRPELC

Aleksis Xenophon The University of Adelaide

No co-movement

1985 1990 1995 2000 2005 20100

5

10

15

20Xinjiang

Agg

rega

te R

GR

PG

row

th R

ate

(%)

Year

1985 1990 1995 2000 2005 20100

10

20

30

40

Ele

ctric

ity C

onsu

mpt

ion

Gro

wth

Rat

e (%

)

RGRPELC

Aleksis Xenophon The University of Adelaide

National level data

Dependent variable∆RGDP ∆ELC

Constant 0.044 -0.006(0.209) (0.894)

ECTt−1 0.033 0.228***(0.597) (0.008)

∆RGDPL1 0.895*** -0.516

(0.001) (0.181)L2 -0.891*** 0.007

(0.006) (0.987)L3 0.034 -0.812**

(0.910) (0.048)∆ELC

L1 -0.076 1.03***(0.652) (0.000)

L2 0.162 -0.682**(0.478) (0.030)

L3 0.214 0.973***(0.288) (0.000)

P-values in parenthesesAleksis Xenophon The University of Adelaide

Signs

RGRPt = β0 + β1ELCt + µt

µt = RGRPt − β0 − β1ELCt

µt > 0 if RGRPt > β0 + β1ELCt

Restore µt to 0 by ↓ RGRPt OR ↑ ELCt

∆RGRPt = β0 + β1µt−1 + ...

I (−)β1 ⇒↓ ∆RGRPt helps restore equilibrium

∆ELCt = β0 + β1µt−1 + ...

I (+)β1 ⇒↑ ∆ELCt helps restore equilibrium

Aleksis Xenophon The University of Adelaide

Short run Granger causality map

ELC⇒RGRPRGRP⇒ELCRGRP⇔ELCRGRP<ELC

(no lags)

Aleksis Xenophon The University of Adelaide

Aggregation and robustness

Short run Granger causal direction[1] [2] [3] [4] [5] [6] [7] [8]

China

Southern ChinaPower Grid

GuangdongGuangxiGuizhouYunnan

ELC⇒RGRP RGRP⇒ELC RGRP⇔ELC RGRP<ELC (no lags)

Model number and sample period key

[1] [2] [3] [4] [5] [6] [7] [8]

1985-2012 1985-06 1986-07 1987-08 1988-09 1989-10 1990-11 1991-12

Aleksis Xenophon The University of Adelaide

Procedure

1 Calculate optimal lag-length in underlying VAR (SBIC)

2 Check for stationarity (ADF Test)

3 Check for co-integration (Johansen Test)

4 Apply VECM

5 Check for serial correlation (LM Test)

Aleksis Xenophon The University of Adelaide