- Email: [email protected]
Focusing on the right targets: Economic factors driving non-hydro renewable energy transition
Accepted Manuscript Focusing on the right targets: Economic factors driving non-hydro renewable energy transition Boqiang Lin, Oluwasola E. Omoju PII:
S0960-1481(17)30459-7
DOI:
10.1016/j.renene.2017.05.067
Reference:
RENE 8834
To appear in:
Renewable Energy
Received Date: 5 October 2016 Revised Date:
13 May 2017
Accepted Date: 21 May 2017
Please cite this article as: Lin B, Omoju OE, Focusing on the right targets: Economic factors driving nonhydro renewable energy transition, Renewable Energy (2017), doi: 10.1016/j.renene.2017.05.067. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Focusing on the right targets: Economic factors driving non-hydro renewable energy transition
1
5 6 7 8 9
Boqiang Lina, b, *, Oluwasola E. Omojuc a
Collaborative Innovation Center for Energy Economics and Energy Policy, Institute for Studies
RI PT
2 3 4
in Energy Policy, Xiamen University, Xiamen, Fujian 361005, PR China b c
Newhuadu Business School, Minjiang University, Fuzhou, 350108, PR China
China Center for Energy Economics Research (CCEER), School of Economics, Xiamen
University, Xiamen, Fujian 361005, PR China
*Corresponding author: Collaborative Innovation Center for Energy Economics and Energy
11
Policy, Institute for Studies in Energy Policy, Xiamen University, Xiamen, Fujian 361005, PR
12
China; +865922186076; Fax: +865922186075; [email protected], [email protected]
M AN U
SC
10
13
Abstract
15
Reducing fossil fuels use while promoting the adoption of clean and renewable energy sources is
16
crucial for mitigating climate change and attaining the sustainable development goals. This study
17
investigates the driving forces of the share of non-hydroelectricity sources in total electricity
18
generation. Using panel data of forty-six developed and developing countries over 1980-2011
19
and employing panel cointegration estimation techniques, we examine the key factors that
20
influence the share of non-hydro renewable energy sources in the short and long run. The results
21
of the study show that the driving factors have different impacts on the size and share of non-
22
hydro renewable energy; and these impacts are mostly in the long run. Oil price increase and
23
financial development has a significant positive effect on the amount and share of non-hydro
24
renewable electricity generation. The impact of trade openness is ambiguous. Economic
25
development enhances the size of non-hydro renewables but undermines its share in total
26
electricity. Foreign direct investments, ratification of the Kyoto protocol, gross fixed capital
27
formation and resource rent have no significant impact on non-hydro renewable electricity
28
generation. Based on the results, appropriate policy recommendations are proffered.
29
Keywords: Renewable energy, non-hydro electricity generation; climate change mitigation; panel
30
cointegration
AC C
EP
TE D
14
1
ACCEPTED MANUSCRIPT
1.0
Introduction
32
The unanimous adoption of the resolutions of the Paris climate agreement and the United Nations
33
Sustainable Development Goals (SDGs) showed the international commitment to stem climate
34
change and promote sustainable development. The climate agreement resolves to limit global
35
temperature increase to below 2oC above pre-industrial levels and achieve zero net carbon
36
emissions in the second half of this century by reducing the proportion of fossil fuel in global
37
energy use, among other measures. Similarly, the SDGs vigorously recognise and incorporate
38
environmental sustainability as a vital element for achieving global development goals. Climate
39
scientists largely agree that the use of fossil fuel generates greenhouse gas emissions, which is
40
the leading cause of anthropogenic climate change. According to data from the United States
41
Energy Information Administration (EIA) online database, global carbon emissions increased
42
from 18.4 billion metric tonnes in 1980 to 32.7 billion metric tonnes in 2012, indicating an
43
increase of 77.5 percent over this period. Advanced countries such as the OECD members and
44
emerging economies such as Brazil, South Africa, China and India are responsible for such
45
phenomena increase. Thus, addressing the global challenge of climate change and global
46
warming requires drastic reduction of CO2 emissions.
47
One of the key measures aimed at addressing energy-related CO2 emission and mitigating climate
48
change and global warming is increasing the share of renewable energy (or reducing the share of
49
fossil-fuel energy) in the total energy mix. While considerable success has been achieved in this
50
respect in countries and regions like the United States, South Korea, Japan, Costa Rica and the
51
European Union, the level of renewable energy adoption in most developing countries is
52
relatively low. Net renewable electricity generation in the United States and Germany in 2011
53
stood at 527.49 TWh and 126.18 TWh respectively compared to 115.36 TWh for the entire
54
Africa continent. This corresponds to 13 percent, 22 percent and 18 percent of total electricity
55
generation in the United States, Germany and entire Africa respectively in the same period. More
56
interestingly, the development and adoption of renewable energy within developed and
57
developing countries varies considerably. For example, while Sweden had attained 65 percent of
58
its electricity consumption from renewable sources as at 2011, the United Kingdom had only
59
achieved 11 percent in the same period. Similarly, Brazil had achieved 86 percent of its
60
electricity requirement from renewable energy sources, mainly hydro, compared to 21 percent by
61
India as at the same period.
AC C
EP
TE D
M AN U
SC
RI PT
31
2
ACCEPTED MANUSCRIPT
The case of developing countries even requires more attention. IEA [1] states that the aggregate
63
energy-related CO2 emissions of non-Annex I developing countries surpassed those of Annex I
64
industrialised and transition countries (that have ratified the Kyoto Protocol) in 2008, and recent
65
growth in global carbon emissions is exclusively attributed to developing countries. It is therefore
66
expected that pro-poor growth in developing countries will result in a significant rise in energy
67
consumption and CO2 emissions in the near future (Gertler at al. [2]; Jakob et al [3]). Based on
68
these facts, there is need for a comprehensive analysis of the drivers and barriers to renewable
69
energy adoption in order to promote renewable energy use on a level that is needed to mitigate
70
climate change. This is in line with SDSN and IDDRI [4], which posits that deep decarbonisation
71
of the energy system requires both individual and globally coordinated decarbonisation efforts.
72
Numerous studies have been conducted on the econometric analysis of the drivers of and barriers
73
to renewable energy development and consumption in recent years (Brunnschweiler [5]; Dogan
74
[6, 7]; Omri and Nguyen [8]; Peterson [9]; Popp et al [10]; Sadorsky [11]; Salim et al [12]). The
75
main contribution of this paper to the literature is threefold. First, although there have been
76
enormous studies on renewable energy in the field of energy and environmental economics, the
77
majority of these studies focus specifically on developed and industrialised countries such as the
78
OECD countries and emerging countries such as China and India. In contrast, this paper analyses
79
the determinants of renewable energy adoption in a panel of both developed and developing
80
countries from various regions such as Africa, Europe, North and South America, Asia, taking
81
cognisance of differences between countries. The impact of economic and policy factors on
82
renewable energy adoption could differ between developed and developing countries, and may
83
also depend on specific development characteristics of each country.
84
Second, majority of previous papers in the literature study aggregate renewable energy,
85
combining hydro and non-hydro renewable energy. In our opinion, this is inappropriate for three
86
main reasons and may obscure the key factors influencing different types of renewable energy.
87
First, the factors that drive renewable energy from hydro and non-hydro are substantially
88
different. According to Johnstone et al [13], the effects of public policies on renewable energy
89
depend on the specific type of renewable energy. Hydroelectricity depends largely on the
90
availability of river flows while non-hydroelectricity such as solar, wind and geothermal does not
91
need this requirement. Given this stark difference in the technical and economic requirements for
92
hydro and non-hydroelectricity, conducting an econometric analysis of the factors driving
AC C
EP
TE D
M AN U
SC
RI PT
62
3
ACCEPTED MANUSCRIPT
aggregate/overall renewable energy may conceal the impact of economic and policy factors on
94
the specific type of renewable energy. Second, in a number of countries with substantial water
95
resources, hydro energy has already been highly exploited and contributes to electricity
96
generation, partly because it is a cheaper source of electricity production than fossil-based
97
generation. Countries like Brazil, Costa Rica and Nigeria generate a large proportion of their
98
electricity from hydro sources. It is preferable in these countries even without considering climate
99
change, unlike solar and wind renewables which are largely driven by climate change mitigation.
100
As an additional option for hydro energy-oriented countries as well as countries without
101
sufficient water resources, non-hydro renewable energy presents a key option. Third,
102
environmental concerns associated with hydro power plants could undermine the potential of this
103
energy source in the near future. It is recognised that hydro power plants cause disruptions in
104
rivers ecosystem and habitats. However, recent studies have shown that acute water shortages as
105
a result of global warming pose a serious challenge to hydro renewable energy in the future
106
(Schmidt et al [14]; van Vliet et al. [15]). Though non-hydro renewable energy sources such as
107
wind and biomass are also subject to the impacts of climate change, the effects are less compared
108
to hydro energy (Pasicko et al [16]). Hence, our focus on non-hydro renewable energy in this
109
paper. There is need to urgently develop non-hydro renewable energy in order to complement for
110
the potential shortfall that could emanate from the effects of global warming-induced water stress
111
on hydro energy contribution. We contribute to this strand of literature by disaggregating
112
renewable energy into hydro and non-hydro, and focusing our research on non-hydro renewable
113
energy sources.
114
The third contribution of this research to the literature is that it focuses on the share of non-hydro
115
renewable energy in total renewable energy, rather than the size of non-hydro renewable energy.
116
Increase in renewable energy consumption without commensurate reduction in fossil fuel energy
117
consumption will not achieve global climate objectives. Hence, we argue that as the consumption
118
of renewable energy increases, the share of fossil fuel consumption in total energy use should
119
reduce. Thus, our paper studies the SHARE (not the SIZE) of renewable energy in total energy
120
consumption. Most of the previous studies on renewable energy analyse the size/amount of
121
renewable energy produced or consumed (Brunnschweiler [5]; Omri and Nugyen [8]; Rafiq et al.
122
[17]; Salim and Rafiq [18]; Sardosky [11]). But it is the share (not the amount or size) of
123
renewable energy in total energy use that is important for climate change mitigation. Pfeiffer and
AC C
EP
TE D
M AN U
SC
RI PT
93
4
ACCEPTED MANUSCRIPT
Mulder [19] and Popp et al. [10], which are very similar to our study because they focus on non-
125
hydro renewable energy, use non-hydro renewable energy per capita and non-hydro (wind, solar
126
PV, geothermal and waste and biomass) net capacity respectively as the dependent variables. In
127
our opinion, using the share of renewables in total energy use as the indicator of energy transition
128
is more appropriate to determine progress towards decarbonisation of the energy system and
129
climate change mitigation. This is a strong departure from existing studies in the literature and a
130
key novelty of this paper. Therefore, this study centers on the share of non-hydro renewable
131
energy in total energy generation.
132
Based on this background, the objective of this paper is to empirically investigate the
133
determining factors of the share of non-hydro renewable energy using data of forty-six developed
134
and developing countries from different regions.
135
1.1
136
Studies on renewable energy have enjoyed distinguishable pedigree in the literature in recent
137
years. This is because adopting renewable energy sources is recognised as one of the major
138
means of decarbonising the energy system and mitigating climate change (Dogan and Seker [20];
139
Moomaw et al. [21]. Similar to this study, various works have been done to determine the factors
140
that influence renewable energy. Lin et al. [22] investigated the factors influencing renewable
141
energy consumption in China, focusing on the electricity sector. Using time series data from
142
1980 to 2011, Johansen cointegration method and error correction model, the study found that
143
financial and economic development significantly enhances renewable energy in China while
144
trade openness, FDI and lobby of fossil fuel undermine it. In another study focusing on
145
renewable energy in the United States, Sovacool [23] asserts that conventional energy sources
146
have potential impacts on the adoption of renewable energy, as utility firms are oriented towards
147
big conventional power plants. The drivers of renewable energy in European countries were
148
explored by Marques et al. [24]. The analysis employed fixed effect vector decomposition
149
(FEVD) technique on data covering the period 1990-2006; and focused on political,
150
socioeconomic and country-specific factors. Based on the results of the study, the goal of
151
reducing energy dependency encourages renewable energy while traditional energy sources and
152
CO2 emissions are not favourable to renewable energy. The determinants of renewable energy
153
consumption in six emerging economies (Turkey, Indonesia, India, Brazil, Philippines and
154
China) were studied by Salim and Rafiq [18]. The results of the study showed that income is a
SC
M AN U
AC C
EP
TE D
Brief review of the literature
RI PT
124
5
ACCEPTED MANUSCRIPT
key driver of renewables in all six countries. In addition, pollutant emissions significantly
156
influence renewable energy in Indonesia, Brazil, India and China. Sadorsky [11] analysed the
157
relationship between income and renewable energy in emerging economies, and found that
158
income has a significant impact on renewable energy consumption. In another study, Omri and
159
Nguyen [8] analysed the driving factors of renewable energy consumption in 64 countries. The
160
study divided the 64 countries into sub-panels of high, middle and low-income countries, and
161
employed the dynamic GMM estimation technique. Based on the results of the study, it was
162
shown that carbon emissions and trade openness are the major driving factors of renewable
163
energy. On the other hand, oil price has a small negative effect on renewable energy in the global
164
and middle-income panels.
165
In a survey study focusing on greenhouse gas mitigation technology transfer, Peterson [9] found
166
that there is little evidence on the impacts of FDI, ODA, trade and other funding sources.
167
Brunnschweiler [5] investigated the impact of financial sector development on renewable energy
168
in non-OECD countries, and the results showed a positive impact of commercial banking. In a
169
similar study, Glemarec [25] examine the role of finance in off-grid sustainable energy access for
170
the poor. Popp et al [10] assessed the factors that determine renewable energy development in 26
171
OECD countries using data from 1991 to 2004. Based on the analysis, they found that energy
172
security, fossil fuel production, future electricity demand and national renewable energy policies
173
do not have significant impact on renewable energy; while ratification of the Kyoto protocol and
174
deployment of low-carbon substitutes such as nuclear and hydro enhance renewable energy. The
175
relationship among renewable energy, GDP and energy prices in OECD countries was
176
investigated by Chang et al. [26]. They found that countries with higher GDP have the capacity
177
to adopt renewable energy despite their high prices. Rafiq et al. [17] analysed and compared the
178
relationship between renewable energy generation, income and CO2 emission in China and India
179
using multivariate vector error correction model.
180
Some studies have gone farther by analysing the factors that promote or hinder specific types of
181
renewable energy. The driving factors and effectiveness of policies to promote wind energy in
182
the United States are analysed by Bird et al. [27] and Menz and Vachon [28]. Huang et al. [29]
183
explore the barriers to wind energy. They find that apart from intermittency of wind energy, other
184
challenges include lack of national policies, poor institutional frameworks, and inadequate
185
economic assessments. In the analysis of non-hydro renewable energy in 108 countries, Pfeiffer
AC C
EP
TE D
M AN U
SC
RI PT
155
6
ACCEPTED MANUSCRIPT
and Mulder [19] found that income, education, democracy and implementation of economic and
187
regulatory instruments have significant positive effect on renewable energy diffusion. Garcia et
188
al. [30] use expert opinion analysis to investigate the potentials of renewable hydrogen storage
189
systems in Europe. Also studying the impact of policy risk on the development of solar PV in
190
Europe, Luthi and Wustenhagen [31] conduct a preference survey among European PV projects
191
developers and find that policy risks play an important role in investment decisions. The
192
summary of the current literature on the determinants of renewable energy use is shown in table
193
1.
AC C
EP
TE D
M AN U
SC
RI PT
186
7
ACCEPTED MANUSCRIPT
194
Table 1: Summary of current literature on the determinants of renewable energy use Scope/Coverage China
Method/Techniques Johansen’s cointegration and error correction model
Sovacool (2009)
United States
Interviews
Marques et al (2010)
European countries
Salim and Rafiq (2012)
Turkey, Indonesia, India, Brazil, China, Philippines 18 emerging countries 64 countries
Fixed effect vector decomposition (FEVD) technique Fully modified OLS (FMOLS), Dynamic OLS and Granger causality Panel cointegration and VECM Dynamic GMM
Peterson (2007)
Global
Literature survey
Brunnschweiler (2010)
119 Non-OECD countries
Panel data estimations
RE production, financial sector variables
Panel data analysis – pooled regression and fixed effects model
RE capacity, global knowledge stock, GDP per capita, growth rate of electricity consumption, Kyoto protocol, production of alternative energy sources
Popp et al (2011) 26 OECD countries
SC
M AN U
Share of RE in total energy use, political, socio-economic and country-specific factors RE consumption, real GDP, oil prices, CO2 emissions
TE D
RE consumption, income, electricity prices Amount of RE, CO2 emissions, oil price, trade openness, per capita GDP RE and variables used in other studies
EP
Omri and Nguyen (2014)
AC C
Sadorsky (2009)
Variables Used Share of renewable electricity in total electricity consumption, financial development, GDP, trade openness, FDI and fossil fuel “lobby effect” Wind energy
Findings Financial development and GDP enhances RE while trade openness, FDI and fossil fuel “lobby effect” undermine it
RI PT
Author(s) Lin et al (2016)
8
Socio-technical factors are responsible for neglect of RE in the US Goal of energy independence enhances RE while traditional energy sources and CO2 emissions impede RE Income have significant impact on RE in all the countries while pollutant emission affect RE in Brazil, China, India and Indonesia Increase in real income have significant positive effect on RE Trade openness and increase in CO2 emissions are the major drivers of RE Little evidence on the impacts of FDI, ODA, trade and other funding sources on greenhouse gas mitigation technology Adoption of Kyoto protocol and commercial banking has a positive influence on RE Ratification of the Kyoto protocol and deployment of nuclear and hydro enhances RE while energy security, fossil fuel production, future electricity demand and national RE policies do not have effect.
ACCEPTED MANUSCRIPT
OECD countries
Panel threshold regression model
Rafiq et al. (2014)
China and India
Multivariate VECM RE, GDP, CO2 emissions
Bird et al. (2005)
12 major wind power producing states in the US
Literature survey and case studies
Menz and Vachon (2006)
39 states in the US
Pooled OLS
Pfeiffer and Mulder (2013)
108 developing countries
Two-stage estimation method
Aguirre and Ibikunle (2014)
EU, OECD and BRICS countries
FEVD and Panel corrected standard errors (PCSE)
Source: Authors’ tabulation
M AN U
SC
Wind power, policy factors such as renewable portfolio standards (RPS), financial incentives, demand for green power, natural gas price volatility, wholesale market rules Wind power, state-level policies and retail choices
EP
TE D
Non-hydro renewable energy per capita, income, education, democracy, trade openness, aid, electricity consumption growth, FDI, energy mix, economic and regulatory instruments Share of RE in total energy use, political factors, socio-economic factors, country-specific factors
AC C
195
Contribution of RE to energy supply, GDP and CPI
9
Countries with higher GDP have the capacity to adopt RE despite their higher prices. The results are mixed for both India and China, and they differ in the short and long run Numerous drivers of wind RE influence one another’s effectiveness, but RPS is the most effective of state policy drivers.
RI PT
Chang et al. (2009)
Positive relationship between RPS and wind power and negative effect of retail choice Regulatory instruments, income, schooling and democracy promote RE while openness, aid, electrification and fossil fuel abundance delay the diffusion of RE.
CO2 emissions, high biomass and solar energy potential enhances RE; energy use, high electricity rates for the industrial sector, fossil fuel use have negative impacts on RE
ACCEPTED MANUSCRIPT
From the review of the current literature on the determinants and drivers of renewable energy,
197
some limitations are identified, which we aim to address in this paper. First, previous studies
198
focus extensively on aggregate renewable energy, with only few distinguishing between hydro
199
and non-hydro. We address this gap by focusing on non-hydro renewable energy. Second, most
200
of the papers use the aggregate amount/size of renewable energy in their analysis. We analyse the
201
driving factors of the share of non-hydro renewable energy in total energy use and argue that
202
economic and policy factors have different impacts on the size and share of renewable energy
203
consumption. Third, unlike previous studies that focus on developed and emerging economies,
204
the sample in this study consists of a panel of developed and developing countries from various
205
regions, subject to data availability. The inclusion of developed and developing countries in the
206
sample is to determine the universal factors that drive the share of non-hydro renewable energy.
207
The contents of the paper are as follows: Section one presents the background of the study and
208
summarises the relevant literature. The method of analysis is the focus of section two. Section
209
three presents the results of the empirical analysis while chapter four consists of the discussion of
210
results, policy implications and conclusion of the study.
211
2.0
Methodology
212
2.1
Variables, empirical model and data sources
213
The empirical analysis of this paper utilises panel data of forty-six developed and developing
214
countries over the period 1980-2011. The countries included in the study are listed in table 2
215
below. The variables included in this paper are based on theoretical understanding and from the
216
literature, subject to data availability. The dependent variable is the share of non-hydroelectricity
217
in total electricity generation. To confirm that there may be differences in the factors that
218
influence the share of non-hydro renewable electricity in total electricity generation and the
219
factors that influence the size of non-hydro electricity generation, we also use the latter as a
220
dependent variable in another model for comparison purpose.
221
Table 2: List of countries in the study Argentina Australia
Austria
Belize
Bolivia
Brazil
Canada
Chile
China
Colombia
Costa Rica
Denmark
Dominican Republic
Fiji
Finland
France
AC C
EP
TE D
M AN U
SC
RI PT
196
10
ACCEPTED MANUSCRIPT
Greece
Guatemala
India
Indonesia
Ireland
Italy
Jamaica
Japan
Jordan
Kenya
Mexico
Netherland
Nicaragua
Norway
New Zealand
Panama
Peru
Philippines
Singapore
Spain
Sweden
Thailand
Trinidad & Tobago
Turkey
Uruguay
United States of America
RI PT
Gabon
Senegal
Switzerland
United Kingdom
Source: Authors’ tabulation
223
To select the independent variables in this paper, we follow the general insights from the energy
224
economics literature. The independent variables in this study include foreign direct investment,
225
trade
226
commitment/institutions. Following Brunnschweiler [5], we include additional regressors - gross
227
fixed capital formation and natural resource dependence in the model - to check for the
228
robustness of the estimates of our variables of interest. Trade openness and foreign direct
229
investment are included in the model because trade linkages and foreign direct investment are
230
expected to enhance technology and knowledge transfer, and consequently facilitate the
231
development of renewable energy technologies (Omri and Nguyen [8]; Pfeiffer and Mulder [19]).
232
Trade openness is calculated as the share of import and export trade in GDP while foreign direct
233
investment is represented by the ratio of foreign direct investment to GDP. GDP per capita
234
represents level of development and capacity to invest in renewable energy technologies. It is
235
expected that as the level of income increases, citizens may begin to demand for more
236
environmental protection (Sardosky [11]). More importantly, improvement in the level of
237
development may enhance a country’s capability to invest in renewable energy and promote
238
sustainable development (Marques et al. [24]). This variable is measured by real GDP per capita.
239
Financial development is supposed to facilitate the development of renewable energy by
240
allocating financial resources to renewable energy investment and projects (Brunnschweiler [5]).
241
This is particularly because of the high investment costs of most renewable energy technologies.
242
Financial development in this study is indicated by domestic credit to the private sector as a
243
percent of GDP. Oil price indicates the price of substitutes. In the standard theory of demand and
per
M AN U
GDP
capita,
financial
development,
oil
price,
policy
AC C
EP
TE D
openness,
SC
222
11
ACCEPTED MANUSCRIPT
supply, the price of a substitute has impact on the demand or supply of a commodity. In this case,
245
we assume that conventional electricity sources from fossil fuel are substitutes for renewable
246
energy (Salim and Rafiq [18]). Thus, an increase in oil price will reduce the demand for oil and
247
fossil-fuel electricity generation and possibly increase the demand for renewable energy. Oil
248
price is the same for all countries but changes over time, so it is taken as a time-specific fixed
249
effect. So we hypothesise that an increase in oil price will lead to increase in renewable energy
250
generation. Oil price is proxied by West Texas Intermediate (WTI) spot prices. The Kyoto
251
protocol is used as a measure of institutional support and commitment to climate policy (Aguirre
252
and Ibikunle [32]; Pfeiffer and Mulder [19]). Like Aguirre and Ibikunle [32] and Popp et al. [10],
253
we adopt a dummy variable to capture it. The value of 1 represents ratification of the protocol
254
and 0 represents no ratification of the protocol. We expect a positive relationship between
255
ratification of the Kyoto protocol and renewable energy development.
256
Several literatures have shown that the more a country depends on fossil fuel, the less it is likely
257
to support the development of renewable energy. This is known as the “lobby effect” (Lin et al.
258
[22]; Marques at al. [24]). In past studies, the share of fossil fuel in total energy consumption is
259
used to capture the lobby effect. However, in this study we indicate it by the share of natural
260
resource/oil rent in GDP. This comprehensively captures the impact of natural resource/oil
261
dependence on renewable energy generation. The idea is to investigate if countries that depend
262
on natural resources, especially oil, see renewable energy adoption as a threat to the potentials of
263
their oil resources. Gross fixed capita formation indicates the level of existing physical
264
infrastructure and assets of a country. Countries with developed physical infrastructural assets
265
may be able to develop renewable energy better and faster than countries with lower physical
266
infrastructure. It is indicated as total gross fixed capital formation as a percentage of GDP. We
267
hypothesise a positive association between all the independent variables (except “lobby effect”)
268
and share of non-hydro renewable electricity generation. Table 3 below shows the variables, their
269
definitions and units of measurement.
270
AC C
EP
TE D
M AN U
SC
RI PT
244
Table 3: Definition of the variables used in the study Variable Definition SNHE Share of non-hydro power in total electricity generation NHE Amount/size of non-hydro power GDPPC GDP per capita 12
Units of measurement Percent TWh US$
ACCEPTED MANUSCRIPT
Percent Percent Percent Percent Percent US$ 0 or 1
RI PT
271
TO Trade openness (as % of GDP) FDI Foreign direct investment (as % of GDP) FIN Credit to the private sector (as % of GDP) GFCF Gross fixed capital formation (as % of GDP) RENT Share of resource/oil rents in GDP OIL WTI spot oil price KYOTO Ratification of the Kyoto Protocol (Dummy variable) Source: Authors’ tabulation
Thus, to investigate the dynamic relationship between the share of non-hydro renewable
273
electricity and its influencing factors, we assume a fixed effect specification following Kao et al.
274
[33] and specify the following baseline model:
SC
272
snheit = αgdppcit + δtoit + βfdiit + ρfinit + γoilt + ηkyotoit + εit………..1
276
To compare the impact of these variables on the share of non-hydro RE and the size of non-hydro
277
RE, we also estimate the following model:
M AN U
275
278
nheit = αgdppcit + δtoit + βfdiit + ρfinit + γoilt + ηkyotoit + εit………….2
279
To check for the robustness of the estimates, we follow Brunnschweiler [5] and include two
280
additional variables (gfcf and rent) to equation 1 and 2 to estimate the following model: snheit = αgdppcit + δtoit + βfdiit + ρfinit + γoilt + ηkyotoit + λgfcfit + φrentit + εit………..3.
282
Equation 1, 2 and 3 provide the mathematical expression of the model that describes the
283
relationship between non-hydro renewable energy and the selected variables. All the variables are
284
as described above in table 3. i refers to the forty-six countries included in the study; t refers to
285
the time period of the study (1980-2011); and εt is the error terms. The parameters in equation 1,
286
2 and 3 (α, δ, β, ρ, γ, η, λ, φ) are the coefficients of the corresponding variables. In other words,
287
they represent the effect of the corresponding variables on the dependent variable (nhe or shne).
288
For example, α, δ, β, ρ, γ, η, λ, φ are the respective estimated effects of GDP per capita, trade
289
openness, foreign direct investment, financial development, oil price, Kyoto protocol, gross fixed
290
capital formation and resource rent on the size or share of non-hydro renewable energy. If the
291
parameters are positive, it indicates that an increase in the value of the corresponding variables
292
has a positive effect on non-hydro renewable energy and vice-versa. The error term εt represents
293
other factors that affect non-hydro renewable energy but not included in the model.
294
The data used in the paper are obtained from various sources. SNHE is calculated based on
295
available data from the database of the United States Energy Information Administration (EIA).
296
Data on non-hydro renewable electricity and total electricity generation are obtained from this
AC C
EP
TE D
281
13
ACCEPTED MANUSCRIPT
database and the share of non-hydro renewable electricity is calculated. Data on GDP per capita,
298
trade openness, foreign direct investment, financial development, gross fixed capital formation
299
and natural resource rents are obtained from the World Development Indicator database of the
300
World Bank. Data on oil price is obtained from BP Statistical Review (BP [34]) while
301
information on the ratification, including date, of the Kyoto protocol is obtained from the United
302
Nations Framework Convention on Climate Change (UNFCCC [35]).
303
2.2
304
There are various techniques for estimating the coefficients of panel data models. The most
305
commonly used methods are the pooled OLS, fixed and random effect models estimators. But
306
these techniques are biased and produce inconsistent estimates when the variables are
307
cointegrated. This leads to the development of new estimators that estimate the cointegration
308
vectors of panel data. The new estimators include within and between-group estimators such as
309
OLS estimators, fully modified OLS (FMOLS) estimators and dynamic OLS (DOLS) estimators
310
(McCoskey and Kao [36]; Phillips and Moon [37]). They produce asymptotically unbiased,
311
normally distributed coefficient estimates (Pedroni [38]; Kao and Chiang [39]). The OLS and
312
DOLS are parametric methods, with the DOLS including the lagged first difference. On the other
313
hand, the FMOLS is a non-parametric method for correcting endogeneity (Chaiboonsri et al.
314
[40]). The OLS is biased and inconsistent when applied to cointegrated panels; but the FMOLS
315
and DOLS are more appropriate for heterogeneous cointegrated panels (Hamit-Haggar [41]). The
316
advantages of these estimation techniques are that they generates consistent estimates of the
317
parameters and also addresses the problems of endogeneity in the cointegration regression and
318
serial correlation in the residual. While Pedroni [38] find that DOLS has higher size distortions
319
than FMOLS, Kao and Chiang [39] show that FMOLS may be more biased than DOLS (Harris
320
and Sollis [42]). Ramirez [43] on the other hand argued that the FMOLS method is preferred
321
over the DOLS method for relatively small samples. In this study, we apply both the FMOLS and
322
DOLS techniques to determine the long run determinants of the share of non-hydro renewable
323
energy, following the procedure stated in box/stage 3 of the data flow diagram in figure 1.
324
We use the dynamic fixed effect estimator (DFE) to derive the short run dynamics of the
325
relationship among the variables, within the context of the error correction model. This technique
326
is based on recent developments in the literature on non-stationary cointegrated panel
RI PT
297
AC C
EP
TE D
M AN U
SC
Estimation technique
14
ACCEPTED MANUSCRIPT
econometrics (Blackburne and Frank [44]). It restricts the coefficient of the cointegrating vector,
328
the speed of adjustment coefficient and the short-run coefficient to be equal across the panels.
329
The error correction model distinguishes between the short and long run relationships among the
330
variables; and the coefficient of the error correction term indicates the speed of adjustment to
331
long run relationships. This procedure is implemented following box/stage 4 of the data flow
332
diagram. For more information on the FMOLS/DOLS, see Kao, Chiang and Chen [33]. See
333
Blackburne and Frank [44], Pesaran, Shin and Smith ([45], [46]) and see Pesaran and Smith [47]
334
for more information on the dynamic fixed effect (DFE).
335
Before carrying out the estimation, there is need to test for the stationarity of the variables as well
336
as the existence of a long run relationship between non-hydro renewable energy and its
337
influencing factors. According to Deng [48], most series of economic variables are non-
338
stationary and using such economic series for analysis would produce unreliable and biased
339
estimates. If the series contain unit roots, their first difference will be taken to eliminate the unit
340
root and make it stationary. Two categories of unit roots tests have been developed for panel
341
data. The first category is the tests that assume that the different cross-section sequences have a
342
common unit root process; and includes Levin-Lin-Chu (LLC) test (Levin et al [49]), Hadri LM
343
test (Hadri [50]) and Breitung test (Breitung [51]). On the other hand, the second category
344
assumes that the cross-section sequences have different individual unit root process; and they
345
include Im-Pesaran-Shin (IPS) test (Im et al [52]), Fisher-ADF and Fisher-PP tests (Choi [53]).
346
Because of their popularity and wide use in the literature, and given the heterogeneity among the
347
countries in this study, the second category tests are employed in this paper and the procedure is
348
shown in box/stage 1 of the data flow diagram. Furthermore, the cointegration test examines the
349
presence of a long run relationship among the variables. In this study, following the procedure in
350
box/stage 2 of the data flow diagram, we use three recently developed panel cointegration tests,
351
which includes Pedroni residual cointegration test (Pedroni [54]), Kao residual cointegration test
352
(Kao [55]) and Fisher-Johansen cointegration test (Maddala and Wu [56]) to examine the
353
existence of long run relationship between non-hydro renewable energy and its determinants.
354
The data flow diagram which presents an overview and steps of the methodological approach is
355
provided below in figure 1. We first conduct the panel unit root rests, followed by panel
356
cointegration test. After establishing a long run relationship, we use the FMOLS/DOLS to
AC C
EP
TE D
M AN U
SC
RI PT
327
15
ACCEPTED MANUSCRIPT
estimate the coefficient of the relationship and further use the DFE to estimate the short run
358
effect. The analysis is implemented using STATA and Eviews softwares. The data used for the
359
analysis in the various stages of the flow diagram is included as a supplementary information.
M AN U
SC
RI PT
357
360 361
Fig. 1: Data flow diagram that shows the steps and stages taken in the methodological approach.
362
3.0
363
The results of the analysis are presented and discussed in this section. This consists of the results
364
of the unit root tests, panel cointegration tests, pooled OLS tests, FMOLS and DOLS tests and
365
the robustness checks, following the sequence of the data flow diagram. This section further
366
discusses the result of the analysis and the findings.
367
3.1
368
The results of the panel unit root tests adopted are shown in table 4 below. Based on the result,
369
most of the variables are non-stationary at levels, but the first differences of all the variables are
370
stationary at 1% significance level. At levels, only FDI, GFCF and RENT are stationary because
371
the IPS, ADF and PP statistics for FDI, GFCF and RENT are significant at either 1, 5 or 10%
372
significance levels. At levels, there is existence of unit root for NHE, SHNE, GDPPC, TO, FIN,
373
OIL as the statistical values of the IPS, Fisher-ADF and PP tests are not significant (their
374
probability values are higher than 1, 5 and 10%). However, after taking the first difference of all
375
the variables, the statistics for all the variables are significant at 1% level. In other words, the
376
probability values of the statistics are less than 1%. This is consistent for the three unit root
EP
TE D
Results
AC C
Panel unit root tests
16
ACCEPTED MANUSCRIPT
377
testing methods used. The unit root test is not conducted for Kyoto because it is a dummy
378
variable that takes 0 and 1.
379
380 381
Table 4: Panel unit root tests Order of Variables IPS Fisher-ADF integration NHE 16.56088 9.6288 SNHE 12.0195 5.7474 GDPPC 8.6089 7.1174 TO 1.8871 2.3725 @levels FDI -2.7078*** -2.4227 FIN 3.1419 3.4297 OIL 17.4021 16.7592 GFCF -4.8136*** -4.8858*** RENT -6.0360*** -6.0410*** NHE -4.4904*** -3.6936*** SNHE -3.9407*** -5.7720*** GDPPC -7.6173*** -8.1801*** TO -13.9811*** -15.2547*** @first difference FDI -17.2587*** -18.6606*** FIN -10.2986*** -11.1284*** OIL -5.0961*** -5.20277*** GFCF -14.8877*** -16.2633*** RENT -12.2308*** -13.2823*** Source: Authors’ tabulation (results extracted from STATA software); Note: *** denotes 1% significance level.
382
3.2
383
The cointegration tests examine the existence of a long run relationships among the variables.
384
Given that all the variables are stationary after first differencing as shown in table 4, we test for
385
the existence of long run relationship among the variables using Pedroni, Kao and Johansen-
386
Fisher cointegration tests.
Fisher-PP
EP
AC C
387
Cointegration tests
TE D
M AN U
SC
RI PT
7.2648 1.7307 9.4946 1.6906 -10.5629*** 2.0277 15.3358 -6.5708*** -4.9371*** -19.5542*** -22.8941*** -16.2047*** -31.4283*** -42.0410*** -26.3115*** -41.0521*** -30.1946*** -37.1959***
Table 5: Pedroni cointegration test H0: No cointegration Panel v statistic Panel v weighted statistic Panel rho-statistic Panel rho weighted statistic Panel PP statistic Panel PP weighted statistic Panel ADF statistic Panel ADF weighted statistic Group rho statistic 17
Statistic -1.8883 -1.1408 2.2074 3.1393 -3.0500*** -1.6275* -3.4653*** -1.8760** 4.8264
p-value 0.97005 0.8730 0.9864 0.9992 0.0011 0.0518 0.0003 0.0303 1.0000
ACCEPTED MANUSCRIPT
388
Group PP statistic -2.0543** 0.0200 Group ADF statistic -2.6169*** 0.0044 Source: Authors’ tabulation (results extracted from Eviews software);
389
Note: *, **, *** denotes 10%, 5% and 1% significance level respectively The result of the Pedroni test in table 5 shows that the null hypothesis (H0) of no cointegration is
391
rejected for the panel PP, panel PP weighted, panel ADF, panel ADF weighted, group PP and
392
group ADF statistics. This is because the probability values associated with the statistics are less
393
than either 1, 5 or 10%. Out of the eleven statistics computed by the Pedroni test, six statistics
394
confirm the existence of panel cointegration among the variables as the statistics are significant
395
at either 1, 5 or 10% significance levels. According to Pedroni [57], the Panel ADF and Group
396
ADF are the most important statistics and should be the yardstick when the results are
397
conflicting.
400
SC
M AN U
399
Table 6: Kao cointegration test H0: No cointegration t-statistic p-value ADF -2.205965** 0.0137 Residual variance 2.806904 HAC variance 2.611304 Source: Authors’ tabulation (results extracted from Eviews software); Note: ** denotes 5% sig. level of the statistic
TE D
398
RI PT
390
The Kao residual and Johansen-Fisher panel cointegration tests are also conducted to confirm the
402
existence of cointegration in the variable. Based on the results of the Kao test (ADF statistics = -
403
2.2060, p = 0.0137) in table 6, the null hypothesis of no cointegration is rejected as the p-value
404
associated with the t-statistic is less than the conventional 5% significance level.
405
Table 7: Johansen Fisher panel cointegration test Trend No intercept and trend Intercept only scenarios Hypothesized Fisher stat Fisher stat Fisher stat Fisher stat no. of CE(s) (from trace (from max- (from trace (from maxstatistics) eigen statistics) eigen statistics) statistics) None 1504.0*** 936.7*** 1623.0*** 1099.0*** At most 1 789.2*** 397.7*** 856.7*** 456.6*** At most 2 464.1*** 232.6*** 468.8*** 256.9*** At most 3 296.2*** 161.2*** 264.2*** 161.1*** At most 4 199.3*** 135.4*** 158.6*** 110.1* At most 5 141.3*** 132.0*** 110.4* 98.19 At most 6 88.80 88.80 107.1 107.1
AC C
EP
401
18
Intercept and trend Fisher stat (from trace statistics) 2055.0*** 1177.0*** 622.0*** 353.5**8 216.9*** 149.3*** 98.58
Fisher stat (from maxeigen statistics) 1085.0*** 584.6*** 317.7*** 181.6*** 119.6** 115.9** 98.58
ACCEPTED MANUSCRIPT
Source: Authors’ tabulation (results extracted from Eviews software);
407
Note: *, **, *** denotes 10%, 5% and 1% sig. level respectively
408
The Johansen-Fisher cointegration test uses the trace statistics and maximum eigenvalues to test
409
the cointegration ranks among the panel series. The results of the Johansen-Fisher test in table 7
410
also show the existence of cointegration as the null hypothesis of no cointegration (none) is
411
rejected at 1% significance level under all the trend deterministic scenarios (no intercept and
412
trend, intercept only, and intercept and trend). The scenarios are included to check if there is
413
difference in the result if the model specification changes. The results of the three panel
414
cointegration tests are largely consistent and provide evidence of a long run relationship among
415
the variables.
416
3.3
417
The motivation for using the FMOLS/DOLS to obtain the long run estimates of the impacts of
418
the selected variables on the share of non-hydro renewable energy in this study is that they
419
address the problems of endogeneity and serial correlation in the model. Using conventional
420
pooled OLS to analyse the determinants of non-hydro renewable power produces unreliable
421
estimates. This is confirmed by the pooled OLS results in table 8 as it shows the presence of
422
serial correlation and heteroscedasticity. The Wooldridge test for autocorrelation (Drukker [58];
423
Wooldridge [59]) shows there is evidence of serial correlation as the null hypothesis of “no first
424
order autocorrelation” is rejected. Similarly, the null hypothesis of “constant variance” in the
425
Breusch-Pagan/Cook-Weisberg test is rejected. More so, there is reverse causality between the
426
dependent and independent variables in the model, which indicates the problem of endogeneity.
427
The results of the FMOLS and DOLS regressions are also presented in table 8. The results on the
428
left side of the table show the impacts of the variables on the share of non-hydro renewables in
429
total electricity generation (snhe) while the right hand side shows the impacts of the variables on
430
the size of non-hydro renewables (nhe). The results of both FMOLS and DOLS techniques,
431
particularly with respect to the sign, size and significance level, are very similar except that the
432
DOLS model has higher adjusted R-squared and lower standard errors and long run variance.
433
Table 8: FMOLS and DOLS regressions snhe – share of non-hydro RE Variables Pooled OLS
SC
RI PT
406
AC C
EP
TE D
M AN U
Long run effects: Fully Modified and Dynamic OLS
19
snhe – share of nonhydro RE
nhe – size of nonhydro RE
DOLS
DOLS
FMOLS
FMOLS
ACCEPTED MANUSCRIPT
-0.110 0.048*** 5.138*** -0.459 0.045*** 0.008 0.983 1.886 3.355
-0.192* 0.080*** 5.854** -2.880* 0.090*** -0.026 0.905 4.513 246.928
435
Note: *, **, *** denotes 10%, 5% and 1% sig. level respectively
436
From table 8, the results show that FDI has a negative but insignificant effect on non-hydro
437
renewable electricity generation. Specifically, FDI has a negative relationship with both the size
438
and share of non-hydro renewable energy, but the effect is insignificant in the long run. In other
439
words, FDI does not promote non-hydro renewable energy. On the contrary, the impact of
440
financial development on the size and share of non-hydro renewable electricity is positive and
441
significant, implying that financial development promotes non-hydro renewable energy. A 1%
442
increase in financial development leads to a 0.06% increase in the share of non-hydro renewable
443
energy and a 0.05-0.08% in the size of non-hydro renewable energy. GDP per capita, which
444
indicates the level of economic development, has a negative effect on the share of non-hydro
445
renewable electricity but has a positive impact on the amount of non-hydro renewable electricity
446
generation. A 1% increase in GDP per capita leads to a 5.14-5.85% increase in the size of non-
447
hydro renewable energy, but leads to a 3.22-2.87% decrease in the share of non-hydro renewable
448
energy. The impact of the Kyoto protocol on non-hydro renewable energy is inconsistent and
449
mostly insignificant. According to a priori expectation, oil price has a positive significant effect
450
on both the share and size of non-hydro renewable energy, indicating the substitutability between
451
fossil fuel and renewable energy. A 1% increase in oil price leads to a 0.05-0.09% increase in the
452
size of non-hydro renewable energy and a 0.02-0.03% increase in the share of non-hydro
453
renewable energy. Trade openness has a significant positive effect on the share of non-hydro
454
renewable electricity in total electricity generation but has an insignificant effect on the level of
AC C
EP
TE D
M AN U
SC
RI PT
434
FDI -0.060 0.031 -0.057 FIN 0.016*** 0.055*** 0.058*** GDPPC -1.269*** -3.222** -2.870*** KYOTO 0.846 1.395** 0.314 OIL 0.0037*** 0.030*** 0.021* TO 0.015*** 0.058*** 0.070*** 2 Adj. R 0.090 0.912 0.781 Standard error 6.792 2.112 3.333 Long run variance 4.639 39.978 Wooldridge serial corr. test 36.630*** Breusch-Godfrey/Cook375.64*** Weisberg Source: Authors’ tabulation (results extracted from Eviews software);
20
ACCEPTED MANUSCRIPT
455
non-hydro renewables. A 1% increase in trade openness leads to a 0.06-0.07% increase in the
456
share of non-hydro renewable energy.
457
The results of the robustness checks are presented in table 9. Two additional independent
458
variables – gfcf and rent - are added to the model to determine if the estimates of the initial
459
variables (fdi, fin, gdppc, kyoto, oil and to) changes significantly in terms of sign, size or level of
460
significance.
RI PT
3.4 Robustness checks
462
Table 9: Robustness checks of the FMOLS and DOLS estimates snhe – share of non-hydro nhe – size of non-hydro RE RE Variables DOLS FMOLS DOLS FMOLS FDI -0.043 -0.061 -0.052 -0.215* FIN 0.051*** 0.056*** 0.049*** 0.078*** GDPPC -2.039** -2.857*** 5.968*** 5.494** 0.679 -0.878 -1.266** -2.879* KYOTO OIL 0.026*** 0.024* 0.061*** 0.100*** TO 0.053*** 0.071*** 0.006 -0.025 GFCF -0.038 -0.017 -0.030 0.080 RENT -0.052 -0.0594 -0.180*** -0.147 Adj. R2 0.782 0.781 0.994 0.905 Standard error 3.326 3.328 1.154 4.510 Long run variance 31.655 39.851 0.512 244.885 Source: Authors’ tabulation (results extracted from Eviews software);
463
Note: *, **, *** denotes 10%, 5% and 1% sig. level respectively
TE D
M AN U
SC
461
Comparing the estimates of the initial variables in equation 1 and 2 (table 8) with that of equation
465
3 (table 9), we find that all the coefficient estimates of the variables do not change significantly
466
in terms of size, sign and level of significance. In other words, the inclusion of these additional
467
variables does not affect the estimation results of the baseline model, implying that the estimates
468
are not subject to model specification. The impact of gross fixed capital formation on the size
469
and share of non-hydro renewable electricity generation is not significant. This indicates that
470
there is no significant evidence that current level of physical infrastructural asset enhance the
471
development of non-hydro renewable electricity generation. Also, the evidence of the “lobby
472
effect” is not supported as the estimates are mostly insignificant. This contradicts Lin et al. [22]
473
and Marques et al. [24]. However, it is important to note that unlike in previous studies, we
474
investigate the impact of natural resource dependency on the desire for renewable energy. This
475
expands the idea of the fossil fuel industry “lobby” in the literature to the overall economic
AC C
EP
464
21
ACCEPTED MANUSCRIPT
interest of a country. The aim is to determine whether oil/resource dependent countries see
477
renewable energy as a threat to their economic sustenance. But there is no concrete evidence that
478
natural resource dependence impede the development of non-hydro renewable energy. Countries
479
that depend on natural resources, especially oil, for economic development do not necessarily see
480
renewable energy as a threat to the potential of their natural resources.
481
3.5
482
The result of the short run dynamics is presented in table 10 below.
Short-run dynamics: Panel error correction model
Table 10: Short run dynamics
483
RI PT
476
484
Std. z-statistic P>|z| Error C -0.468 2.189 -0.21 0.831 ∆FDI(-1) 0.009 0.014 0.68 0.499 ∆FIN(-1) -0.004 0.004 -1.23 0.218 ∆GDPPC(-1) -1.611 1.348 -1.19 0.232 ∆KYOTO(-1) -0.236 0.280 -0.84 0.400 ∆OIL(-1) -0.015*** 0.005 -3.04 0.002 ∆TO(-1) -0.001 0.006 -0.11 0.914 ECM(-1) -0.106*** 0.014 -7.70 0.000 Source: Authors’ tabulation (results extracted from Eviews software);
485
Note: *** denotes 1% sig. level
SC
Coefficient
TE D
M AN U
Variable
The result confirms the evidence of a long run relationship among the variables, as the ECM term
487
is significant and negative. The coefficient of the ECM (-0.106), which is the speed of
488
adjustment, implies that if there is a shock to the model, equilibrium will be restored at the speed
489
of 10.6% per year. Among all the variables, only oil price has a significant relationship with non-
490
hydro renewable energy in the short term.
491
4.0
492
The results of the study show that foreign direct investment has an insignificant effect on the
493
share and size of non-hydro renewable electricity generation both in the short and long run, as the
494
coefficients of foreign direct investment in tables 8 and 10 are not significant. This is contrary to
495
expectation but supports recent finding by Peterson [9] that there is little evidence that FDI
496
enhances renewable energy. Lin at al. [22] also finds that FDI undermine renewable energy in
497
China. Pfeiffer and Mulder [19] similarly show that increasing foreign direct investment
498
undermine the adoption and amount of non-hydro renewable energy. On the other hand, Kwakwa
499
[60] finds that foreign direct investment significantly impacts hydro power generation in Ghana.
AC C
Discussion
EP
486
22
ACCEPTED MANUSCRIPT
Foreign direct investment is expected to facilitate technology, skills and knowhow transfer,
501
which are expected to enhance local capacity to promote technology advancements; but there
502
seems to be no direct link between foreign direct investment and renewable energy generation.
503
On the contrary, foreign direct investment may increase overall energy consumption, possibly
504
from fossil fuel energy sources. Potential increase in overall energy consumption may crowd out
505
any increase in renewable energy, thereby undermining the effect on the share of renewable
506
energy in the energy mix. Furthermore, the pollution haven hypothesis postulates that foreign
507
direct investment may cause or aggravate environmental problems in their host countries. The
508
empirical evidence of a positive impact of foreign direct investment on the share of non-hydro
509
renewable electricity in total electricity generation is not strong. This could be also due to the
510
nature and composition of foreign direct investment, as different forms of foreign direct
511
investment may impact renewable energy in different ways.
512
Financial development has a strong positive impact on the amount and share of non-hydro
513
renewable electricity. The coefficient of financial development for both the share and size of
514
non-hydro renewable energy regardless of the estimation technique are positive and significant as
515
shown in table 8. This indicates that financial development contributes to both the level of non-
516
hydro renewable electricity and the share of non-hydro renewable electricity in total electricity
517
generation. This is in line with the findings of Brunnschweiler [5], which show that financial
518
intermediation, particularly commercial banking, has a significant positive effect on renewable
519
energy production, and the impact is large for non-hydropower renewable energy such as wind,
520
solar, geothermal and biomass. But this contrasts with the result of Pfeiffer and Mulder [19]
521
which find that access to finance has no significant effect on non-hydro renewable energy
522
diffusion. Given the high investment costs of renewable energy projects, access to finance is
523
essential to achieve high penetration and production of renewable energy. As the financial sector
524
develops, its capability to fund major long term projects and infrastructure, including renewable
525
energy projects, improves. But investing in renewable energy projects could be considered a risky
526
venture by the financial sector due to high initial cost outlays, long project payback periods as
527
well as risks associated with uncertainty about future global energy and climate change policies
528
(IEA [61]. Hence, the importance of the financial sector to the development of renewable energy
AC C
EP
TE D
M AN U
SC
RI PT
500
23
ACCEPTED MANUSCRIPT
should be recognised and necessary policies should be put in place to mitigate the risks
530
associated with the financing of renewable energy projects.
531
GDP per capita, which indicates the level of economic development, has a significant positive
532
effect on the size of non-hydro renewables and a significant negative effect on the share of non-
533
hydro renewable energy. According to table 8, the coefficient of GDP per capita for size of non-
534
hydro renewable energy is positive but is negative for the share of non-hydro renewable energy.
535
The positive effect of economic development on the size of non-hydro renewable energy in our
536
study corroborates the earlier results of Pfeiffer and Mulder [19]. But economic development
537
reduces the share of renewables in total electricity production. This implies that while economic
538
development may enhance the production of renewable energy, it may adversely affect the share
539
of renewables in total energy production. Economic development increases the capacity to invest
540
in renewable energy and other sustainable development policies and projects. However, it also
541
leads to increase in overall energy/electricity consumption (Saidi and Hammami [62]). This has
542
significant implications for policy. Economic development may increase the amount of non-
543
hydro renewable electricity generated, but in a lesser proportion than it increases overall
544
electricity generation. In other words, development leads to increase in energy use, but the
545
increase is met by increasing the supply of conventional fossil fuel electricity relative to non-
546
hydro renewable electricity. Economic development is expected to increase the capacity to invest
547
in renewable energy generation. However, it could also increase total electricity use relative to
548
the increase in non-hydro renewable electricity. Thus, economic development is not a sufficient
549
prerequisite to facilitate the transition to renewable energy.
550
Based on the coefficient of KYOTO in table 8, the impact of the Kyoto protocol on both the size
551
and share of non-hydro renewable energy is insignificant, implying that the ratification of the
552
Kyoto protocol has no impact on the acceleration of renewable energy. This is also in line with
553
the results of Pfeiffer and Mulder [19], which shows a very weak link between ratification of the
554
protocol and the probability of investing in non-hydro renewable energy. Almer and Winkler [63]
555
also find very little evidence for the effectiveness of the Kyoto Protocol. This might be because
556
the Kyoto protocol focuses less on renewable energy and technology policies in favour of other
557
market-oriented policies. To achieve the goals of the Kyoto protocol, three basic mechanisms -
558
joint implementation, clean development mechanism, and emissions trading - were provided for
AC C
EP
TE D
M AN U
SC
RI PT
529
24
ACCEPTED MANUSCRIPT
and applied by most industrialised countries. In the Kyoto protocol, there was no direct emphasis
560
on technology policies that would enhance the development of renewable energy technology. The
561
protocol’s trading mechanisms are inadequate to enhance renewable energy as argued by Driesen
562
[64], partly because they placed much emphasis on emission reduction through trading.
563
Oil price has a significant positive relationship with non-hydro renewable energy in the long run.
564
The coefficient of OIL in table 8 is positive and significant across methods. This shows that
565
increase in oil price has significant impact on improving both the amount of non-hydro
566
renewable energy production and the share of non-hydro renewables in total energy production.
567
This confirms that conventional fossil-fuel and renewable energy are substitutes in the long run
568
(Pfeiffer and Mulder [19]). Increase in the price of oil reduces the demand for it and enhances the
569
competitiveness and use of renewable energy. But, the impact of oil price increase on the level of
570
non-hydro power is greater than the impact on the share of non-hydro power, signifying that
571
increase in oil price increases the amount of non-hydro renewable power generation but not in
572
the same proportion that it reduces the amount of fossil fuel power generation. This indicates that
573
oil price increase boosts the production of renewable energy but reduces fossil fuel energy
574
demand in a lesser proportion. This underlines the dominant role that fossil fuel plays in the
575
current energy landscape and the possible difficulties that may be associated with phasing it out
576
in the near future.
577
Trade openness does not have significant impact on the size of non-hydro renewables (as the
578
coefficient in table 8 is not statistically significant) but has a significant positive impact on the
579
share of non-hydro renewable in total energy production. With respect to the insignificant effect
580
of trade openness on the size of non-hydro renewable energy, this study is in line with the
581
findings of Pfeiffer and Mulder [19]. On the contrary, Omri and Nguyen [8] find trade openness
582
to be an important driver of the amount of aggregate renewable energy. Trade openness is
583
expected to facilitate the exchange of green technologies and also creates opportunities for
584
human capital accumulation and development. This finding suggests that trade openness may not
585
show appreciable effect on the amount of non-hydro renewable electricity generation, but has a
586
significant impact on the share of non-hydroelectricity in total electricity generation in the long
587
term. This result is counterintuitive and should be taken with some caution. However, a plausible
588
explanation is that the impact of trade openness may depend on the composition and direction of
AC C
EP
TE D
M AN U
SC
RI PT
559
25
ACCEPTED MANUSCRIPT
trade as well as on the type of renewable energy. According to EIA [65], sectoral energy intensity
590
is heavily influenced by the composition of domestic and internationally traded goods and
591
services. In addition, the transmission mechanism of trade openness on renewable energy needs
592
to be comprehensively investigated.
593
From the results of the panel error correction model in table 10, all the variables except oil price
594
do not have a significant impact on non-hydro renewable energy in the short run. Contrary to
595
expectation, oil price has a significant negative relationship with non-hydro renewable energy in
596
the short run, indicating that oil price increase leads to reduction in the share of non-hydro
597
renewables. Marques et al. [24] also reached similar conclusion for EU member countries while
598
Omri and Nguyen [8] find the same results for middle-income countries and in a global panel
599
analysis. This could be as a result of speculation as increase in oil price could trigger oil
600
production relative to the consequent increase in the demand for oil substitutes. Also, in the
601
absence of climate change mitigation policies, increase in the price of oil may lead to the use of
602
alternative fossil fuels such as coal and natural gas rather than renewables, as argued by van
603
Ruijven and van Vuuren [66]. This is even more likely given that increasing the production of
604
renewable energy in response to oil price increase takes time compared to alternative fossil fuel
605
energy sources. The insignificance of most of the variables in the short run indicates that long
606
term planning is required for renewable energy transmission as policies may not show
607
appreciable impacts in the short term. This is obvious given the long time frame for the
608
development of renewable energy projects and infrastructure and the potential challenges
609
associated with phasing out of fossil fuel energy sources.
610
4.1
611
Based on the result of the study, some policy suggestions are provided.
612
•
EP
TE D
M AN U
SC
RI PT
589
AC C
Policy implications
According to the results in table 10, most of the factors, except oil price, have impacts only in
613
the long term. Hence, policy makers should take cognisance of the long term it takes for
614
policies to have significant impacts on renewable energy transition. Countries should develop
615
long term policy mechanisms and integrate renewable energy transmission with long term
616
economic development plans.
617 618
•
The results of the study in table 8 shows that economic factors have differing impacts on the size and share of renewable energy in total energy use. Therefore, policies should be focused
26
ACCEPTED MANUSCRIPT
619
on not only increasing the size of non-hydro renewable energy, but more importantly on
620
increasing the share of non-hydro renewable energy in total energy production. Thus, efforts
621
should be made to promote the factors that increase renewable energy consumption while
622
reducing conventional fossil fuel use. •
Comparing the impact of economic factors on the size and share of non-hydro renewable
RI PT
623
energy in this study with previous studies that analyse aggregate renewable energy, the study
625
shows that economic factors may have different impact on different types of renewable
626
energy. As a result, there should also be emphasis on customising policies for different forms
627
of renewable energy, rather than adopting a one-size-fit-all policy. In other words, policy
628
makers should recognise the potential different impacts of economic factors and policy
629
instruments on different forms of renewable energy and direct targeted policies for the
630
prioritised form of renewable energy. •
M AN U
631
SC
624
Financial development is a key factor promoting both the size and share of non-hydro
632
renewable energy. Hence, development of the financial sector to enhance access to finance
633
for renewable energy investors should be a key priority and policy action for decision makers.
634
•
Increase in the price of oil also increases the share of non-hydro renewable energy in the long term, indicating that renewable energy may serve as a substitute for oil (or fossil fuel) if there
636
is a sustained increase in oil price making renewable energy price to be competitive. Hence,
637
to promote renewable energy, policy makers should eliminate artificial mechanisms like
638
fossil fuel subsidies that keeps oil price down (and by implications reduce the price of fossil-
639
fuel electricity source). •
EP
640
TE D
635
It is also important to note that most of economic and policy factors have long run relationship and affect one another. This is confirmed by the existence of a cointegration
642
relationship in table 5, 6 and 7. This should be taken into consideration when planning and
643
designing policies to facilitate renewable energy transition.
AC C
641
644
4.2
Conclusion
645
This paper investigates the driving factors of non-hydro renewable energy in a panel of forty-six
646
developed and developing countries from various regions, using panel cointegration techniques
647
and error correction model. The results show that trade openness, financial development, foreign
648
direct investment, economic development and ratification of the Kyoto protocol do not have
27
ACCEPTED MANUSCRIPT
impact on non-hydro renewable energy in the short run. In the long run, trade openness enhances
650
the share of non-hydro renewables in total electricity generation. Increase in oil price and
651
financial development have the effect of increasing both the amount and share of non-hydro
652
renewable electricity in total electricity generation. Economic development has a positive
653
significant impact on enhancing the amount of non-hydro renewables but has a negative effect on
654
the share of non-hydro renewables in total electricity generation. This is because economic
655
development may not increase renewable energy in the same or higher proportion than increase
656
in overall energy consumption. On the contrary, ratification of the Kyoto protocol and foreign
657
direct investment does not have significant effect on non-hydro renewable energy in the long run.
658
In summary, the results of this study indicate that economic variables and policies may affect
659
types of renewable energy differently as well as their size and share in total energy use.
660
This paper examines the driving forces of renewable energy transition, focusing on non-hydro
661
renewable energy. Future studies should further examine and possibly compare the impact of
662
these driving forces on both hydro and non-hydro renewable energy. In addition, the impact of
663
the variables could also differ between different sub-types of non-hydro renewable energy such
664
as solar, wind, wave, geothermal, etc. Thus, future research could also analyse the possibility.
665
Future studies also need to investigate the composition and structure as well as the transmission
666
mechanism of the impact of trade openness and foreign direct investment on renewable energy.
667
TE D
M AN U
SC
RI PT
649
Acknowledgments
669
The paper emanates from the co-author’s (O. E. Omoju) research visit to the Institute for
670
Sustainable Economic Development of the University of Natural Resources and Life Sciences
671
(BOKU), Vienna, Austria. The research visit was supported by the Ernst Mach Grant of the
672
Austrian Agency for International Cooperation in Education and Research (OeAD-GmbH). The
673
authors particularly acknowledge the scientific contributions and review comments of the hosts -
674
Prof. Klaus Salhofer, Dr. Ulrich Morawetz and Dr. Johannes Schmidt - and other participants at
675
the Institute’s seminars. We also thank the two anonymous reviewers for their comments and
676
suggestions on the paper.
AC C
EP
668
677 678
28
ACCEPTED MANUSCRIPT
679 680 681
References
683
[1] IEA. CO2 emissions from fuel combustion – Highlights. International Energy Agency, Paris,
684
2010.
685
[2] P. Gertler, O. Shelef, C. Wolfram, A Fuchs, How pro-poor growth affects the demand for
686
energy. NBER Working Paper Series No. 19092, National Bureau of Economic Research,
687
Cambridge, MA, 2013.
688
[3] M. Jakob, M. Haller, R. Marchinski, Will history repeat itself? Economic convergence and
689
convergence in energy use patterns, Energy Economics 34(2011) 95-104.
690
[4] SDSN, IDDRI, Pathways to deep decarbonisation - 2014 interim report. Sustainable
691
Development Solutions Network (SDSN) and Institute for Sustainable Development and
692
International Relations (IDDRI), Columbia University, New York, 2014.
693
[5] C.N. Brunnschweiler, Finance for renewable energy: An empirical analysis of developing and
694
transition economies, Environment and Development Economics 15(2010) 241-274.
695
[6] E. Dogan, Economic growth and electricity consumption from renewable and non-renewable
696
sources: A study of Turkey, Renewable & Sustainable Energy Reviews 52(2015) 534-546.
697
[7] E. Dogan, Analyzing the linkage between renewable and non-renewable energy consumption
698
and economic growth by considering structural break in time-series data, Renewable Energy
699
99(2016) 1126-1136.
700
[8] A. Omri, D.K. Nguyen, On the determinants of renewable energy consumption: International
701
evidence. Energy 72(2014) 554-560.
702
[9] S. Peterson, Greenhouse gas mitigation in developing countries through technology transfer?
703
A survey of empirical evidence, Mitigation and Adaptation Strategies for Global Change
704
13(2007) 283-305.
705
[10] D. Popp, I. Hascic, N. Medhi, Technology and the diffusion of renewable energy, Energy
706
Economics 33(2011) 648-662.
707
[11] P. Sardosky, Renewable energy consumption and income in emerging economies, Energy
708
Policy 37(2009) 4021-4028.
AC C
EP
TE D
M AN U
SC
RI PT
682
29
ACCEPTED MANUSCRIPT
[12] R.A. Salim, K. Hassan, S. Shafiei, Renewable and non-renewable energy consumption and
710
economic activities: Further evidence from OECD countries, Energy Economics 44(2014) 350-
711
360.
712
[13] N. Johnstone, I. Hascic, D. Popp, Renewable energy policies and technological innovation:
713
Evidence based on patent counts, Environmental and Resource Economics 45(2010) 133-155.
714
[14] J. Schmidt, R. Cancella, A.O. Pereira Jnr, The effect of windpower on long-term variability
715
of combined hydro-wind resources: The case of Brazil. Renewable and Sustainable Energy
716
Reviews 55(2016) 131-141.
717
[15] M.T.H. van Vliet, D. Wiberg, S. Leduc, K. Riahi, Power-generation system vulnerability
718
and adaptation to changes in climate and water resources, Nature Climate Change 6(2016) 375-
719
380.
720
[16] R. Pasicko, C. Brakovic, Z. Simic, Assessment of climate change impacts on energy
721
generation from renewable sources in Croatia, Renewable Energy 46(2012) 224-231.
722
[17] S. Rafiq, H. Bloch, R. Salim, Determinants of renewable energy adoption in China and
723
India: A comparative analysis, Applied Economics 46(2014) 2700-2710.
724
[18] R.A. Salim, S. Rafiq, Why do some emerging economies proactively accelerate the adoption
725
of renewable energy, Energy Economics 34(2012) 1051-1057.
726
[19] B. Pfiffer, P. Mulder, Explaining the diffusion of renewable energy technology in
727
developing countries, Energy Economics 40(2013), 285-296.
728
[20] E. Dogan, F. Seker, The influence of real output, renewable and non-renewable energy,
729
trade and financial development on carbon emissions in the top renewable, Renewable &
730
Sustainable Energy Reviews 60(2016) 1074-1085.
731
[21] W. Moomaw, F. Yamba, M. Kamimoto, L. Maurice, J. Nyboer, K. Urama, T. Weir,
732
Introduction, in: O. Edenhofer, R. Pichs-Madruga, Y. Sokona, K. Seyboth, P. Matschoss, S.
733
Kadner, T. Zwickel, P. Eickemeier, G. Hansen, S. Schlomer, C. von Stechow (Eds.), IPCC
734
Special Report on Renewable Energy Sources and Climate Change Mitigation, Cambridge
735
University Press, New York, 2011.
736
[22] B. Lin, O.E. Omoju, U.J. Okonkwo, Factors influencing renewable electricity consumption
737
in China, Renewable & Sustainable Energy Reviews 55(2016) 687-696.
AC C
EP
TE D
M AN U
SC
RI PT
709
30
ACCEPTED MANUSCRIPT
[23] B. Sovacool, Rejecting renewables: The socio-technical impediments to renewable
739
electricity in the United States, Energy Policy 37(2009) 4500-4513.
740
[24] A.C. Marques, J.A. Fuinhas, J.R.P. Manso, Motivations driving renewable energy in
741
European countries: A panel data approach, Energy Policy 38(2010) 6877-6885.
742
[25] Y. Glemarec, Financing off-grid sustainable energy access for the poor. Energy Policy
743
47(2012) 87-93.
744
[26] T.H. Chang, C.M. Huang, M.C. Lee, Threshold effect of the economic growth rate on the
745
renewable energy development from a change in energy price: Evidence from OECD countries,
746
Energy Policy 37(2009) 5796-5802.
747
[27] L. Bird, M. Bolinger, T. Gagliano, R. Wiser, M. Brown, B. Parsons, Policies and market
748
factors driving wind power development in the United States, Energy Policy 33(2005) 1397-
749
1407.
750
[28] F. Menz, S. Vachon, The effectiveness of different policy regimes for promoting wind
751
power: Experience from the States, Energy Policy 34(2006) 1786-1796.
752
[29] S.C. Huang, S.L. Lo, Y.C. Lin, To re-explore the causality between barriers to renewable
753
energy development: A case study of wind energy, Energies 6(2013) 4465-4488.
754
[30] D.A. Gracia, F. Barbanera, F. Cumo, U. Di Matteo, B. Nastasi, Expert opinion analysis on
755
renewable hydrogen storage systems potential in Europe, Energies 9(2016) 1-22.
756
[31] S. Luthi, R. Wustenhagen, The price of policy risk – Empirical insights from choice
757
experiments with European photovoltaic project developers, Energy Economics 34(2012) 1001-
758
1011.
759
[32] M. Aguirre, G. Ibikunle, Determinants of renewable energy growth: A global sample
760
analysis, Energy Policy, 69(2014) 374-384.
761
[33] C. Kao, M.H. Chiang, B. Chen, International R&D spillovers: An application of estimation
762
and inference in panel cointegration, Oxford Bulletin of Economics and Statistics 61(1999) 693-
763
711.
764
[34] BP, BP statistical review of world energy, BP Plc, London, 2016.
765
[35] UNFCCC, Kyoto Protocol - Status of ratification, United Nations Framework Convention
766
on
AC C
EP
TE D
M AN U
SC
RI PT
738
Climate
Change,
31
Bonn,
2009.
ACCEPTED MANUSCRIPT
http://unfccc.int/files/kyoto_protocol/status_of_ratification/application/pdf/kp_ratification.pdf
768
(accessed 23.06.16)
769
[36] S. McCoskey, C. Kao, Comparing panel data cointegration tests with an application of the
770
twin deficits problem, Mimeo, Syracuse University, New York, 1999.
771
[37] P.C.B. Phillips, H. Moon, Linear regression limit theory for non-stationary panel data,
772
Econometrica 67(1999) 1057–1111.
773
[38] P. Pedroni, Fully modified OLS for heterogeneous cointegrated panels: in B.H. Baltagi, T.B.
774
Fomby, R.C. Hill (Eds.), Nonstationary Panels, Panel Cointegration, and Dynamic Panels
775
(Advances in Econometrics, Volume 15), Elsevier Science Inc., 2000, pp. 93-130.
776
[39] C. Kao, M.H. Chiang, On the estimation and inference of a cointegrated regression in panel
777
data, Advances in Econometrics 15(2000) 179-222.
778
[40] C. Chaiboonsri, J. Sriboonjit, T. Sriwichailamphan, P. Chaitip, S. Sriboonchitta, A panel
779
cointegration analysis: An application to international tourism demand of Thailand, Annals of the
780
University of Petrosani, Economics, 10(2010), 69-86.
781
[41] M. Hamit-Haggar, Greenhouse gas emissions, energy consumption and economic growth: A
782
panel cointegration analysis from Canadian industrial sector perspective, Energy Economics
783
34(2012) 358-364.
784
[42] R. Harris, R. Sollis, Applied time series modeling and forecasting. John Wiley & Sons Ltd,
785
England, 2005.
786
[43] M.D. Ramirez, A panel unit root and panel cointegration test of the complementarity
787
hypothesis in the Mexican case: 1960-2001, Atlantic Economic Journal 35(2007) 343-356.
788
[44] E.F. Blackburne, M.W. Frank, Estimation of nonstationary heterogeneous panels, The
789
STATA Journal 7(2007) 197-208.
790
[45] M.H. Pesaran, Y. Shin, R.P. Smith, Estimating long-run relationships in dynamic
791
heterogeneous panels, DAE Working Papers Amalgamated Series No. 9721, 1997.
792
[46] M.H. Pesaran, Y. Shin, R.P. Smith, Pooled mean group estimation of dynamic
793
heterogeneous panels, Journal of the American Statistical Association 94(1999) 621-634.
794
[47] M.H. Pesaran, R.P. Smith, Estimating long-run relationships from dynamic heterogeneous
795
panels, Journal of Econometrics 68(1995) 79-113.
AC C
EP
TE D
M AN U
SC
RI PT
767
32
ACCEPTED MANUSCRIPT
[48] A. Deng, Understanding spurious regression in financial economics, Journal of Financial
797
Econometrics, 12(2014) 122-150.
798
[49] A. Levin, C.F. Lin, C.S.J Chu, Unit root tests in panel data: Asymptotic and finite-sample
799
properties, Journal of Econometrics 108(2002) 1-24.
800
[50] K. Hadri, Testing for stationarity in heterogeneous panel data, Econometrics Journal 3(2000)
801
148-161.
802
[51] J. Breitung, The local power of some unit root tests for panel data, in: B.H. Baltagi (ed.),
803
Nonstationary Panels, Panel Cointegration, and Dynamic Panels: Advances in Econometrics,
804
Volume 15: JAY Press, Amsterdam, 2000, pp. 161-178.
805
[52] K.S. Im, M.H. Pesaran, Y. Shin, Testing for unit roots in heterogeneous panels, Journal of
806
Econometrics 115(2003) 53-74.
807
[53] I. Choi, Unit root tests for panel data, Journal of International Money and Finance 20(2001)
808
249-272.
809
[54] P. Pedroni, Panel cointegration: Asymptotic and finite sample properties of pooled time
810
series tests with an application to the PPP hypothesis, Econometric Theory 20(2004) 597-625.
811
[55] C. Kao, Spurious regression and residual-based tests for cointegration in panel data when
812
the cross-section and time series dimensions are comparable, Journal of Econometrics 90(2004)
813
1-44.
814
[56] G.S Maddala, S. Wu, A comparative study of unit root tests with panel data and a new
815
simple test, Oxford Bulletin of Economics and statistics 61(1999) 631-652.
816
[57] P. Pedroni, Critical values for cointegration tests in heterogeneous panels with multiple
817
regressors, Oxford Bulletin of Economics and Statistics 61(1999), 653-670.
818
[58] D.M. Drukker, Testing for serial correlation in linear panel-data models, Stata Journal
819
3(2003) 1-10.
820
[59] J.M. Wooldridge, Econometric analysis of cross section and panel data. The MIT Press,
821
Cambridge, Massachusetts, 2002.
822
[60] P.A. Kwakwa, An investigation into the determinants of hydropower generation in Ghana,
823
MPRA Working Paper No. 68033, Munich Personal RePEc Archive, Munich, 2015.
824
[61] IEA, 2005. Reducing greenhouse gas emissions: The potential of coal. Coal Industry
825
Advisory Board, International Energy Agency, Paris.
AC C
EP
TE D
M AN U
SC
RI PT
796
33
ACCEPTED MANUSCRIPT
[62] K. Saidi, S. Hammami, The impact of CO2 emissions and economic growth on energy
827
consumption in 58 countries, Energy Reports 1(2015) 62-70.
828
[63] C. Almer, R. Winkler, Analysing the effectiveness of international environmental policies:
829
The case of the Kyoto Protocol, Journal of Environmental Economics and Management 82(2017)
830
125-151.
831
[64] D.M. Driesen, Renewable energy under the Kyoto protocol: The case for mixing
832
instruments. College of Law Faculty Scholarship Paper No. 27, Syracuse University, 2007.
833
[65] EIA, Annual energy outlook 2016 with projections to 2040, Energy Information
834
Administration, Washington, D.C., 2016.
835
[66] B. van Ruijven, D.P. van Vuuren, Oil and natural gas prices and greenhouse gas emission
836
mitigation, Energy Policy 37(2009) 4797-4808.
AC C
EP
TE D
M AN U
SC
RI PT
826
34
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
837
35
ACCEPTED MANUSCRIPT
EP
TE D
M AN U
SC
RI PT
We examine the driving factors of non-hydro renewable energy The study focuses on both the size and share of non-hydro renewable energy Oil price and financial development have impact on non-hydro renewable energy Economic factors impact the size and share of non-hydro renewables in different ways The impacts of economic and policy factors are mostly in the long run
AC C
• • • • •
S0960-1481(17)30459-7
DOI:
10.1016/j.renene.2017.05.067
Reference:
RENE 8834
To appear in:
Renewable Energy
Received Date: 5 October 2016 Revised Date:
13 May 2017
Accepted Date: 21 May 2017
Please cite this article as: Lin B, Omoju OE, Focusing on the right targets: Economic factors driving nonhydro renewable energy transition, Renewable Energy (2017), doi: 10.1016/j.renene.2017.05.067. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Focusing on the right targets: Economic factors driving non-hydro renewable energy transition
1
5 6 7 8 9
Boqiang Lina, b, *, Oluwasola E. Omojuc a
Collaborative Innovation Center for Energy Economics and Energy Policy, Institute for Studies
RI PT
2 3 4
in Energy Policy, Xiamen University, Xiamen, Fujian 361005, PR China b c
Newhuadu Business School, Minjiang University, Fuzhou, 350108, PR China
China Center for Energy Economics Research (CCEER), School of Economics, Xiamen
University, Xiamen, Fujian 361005, PR China
*Corresponding author: Collaborative Innovation Center for Energy Economics and Energy
11
Policy, Institute for Studies in Energy Policy, Xiamen University, Xiamen, Fujian 361005, PR
12
China; +865922186076; Fax: +865922186075; [email protected], [email protected]
M AN U
SC
10
13
Abstract
15
Reducing fossil fuels use while promoting the adoption of clean and renewable energy sources is
16
crucial for mitigating climate change and attaining the sustainable development goals. This study
17
investigates the driving forces of the share of non-hydroelectricity sources in total electricity
18
generation. Using panel data of forty-six developed and developing countries over 1980-2011
19
and employing panel cointegration estimation techniques, we examine the key factors that
20
influence the share of non-hydro renewable energy sources in the short and long run. The results
21
of the study show that the driving factors have different impacts on the size and share of non-
22
hydro renewable energy; and these impacts are mostly in the long run. Oil price increase and
23
financial development has a significant positive effect on the amount and share of non-hydro
24
renewable electricity generation. The impact of trade openness is ambiguous. Economic
25
development enhances the size of non-hydro renewables but undermines its share in total
26
electricity. Foreign direct investments, ratification of the Kyoto protocol, gross fixed capital
27
formation and resource rent have no significant impact on non-hydro renewable electricity
28
generation. Based on the results, appropriate policy recommendations are proffered.
29
Keywords: Renewable energy, non-hydro electricity generation; climate change mitigation; panel
30
cointegration
AC C
EP
TE D
14
1
ACCEPTED MANUSCRIPT
1.0
Introduction
32
The unanimous adoption of the resolutions of the Paris climate agreement and the United Nations
33
Sustainable Development Goals (SDGs) showed the international commitment to stem climate
34
change and promote sustainable development. The climate agreement resolves to limit global
35
temperature increase to below 2oC above pre-industrial levels and achieve zero net carbon
36
emissions in the second half of this century by reducing the proportion of fossil fuel in global
37
energy use, among other measures. Similarly, the SDGs vigorously recognise and incorporate
38
environmental sustainability as a vital element for achieving global development goals. Climate
39
scientists largely agree that the use of fossil fuel generates greenhouse gas emissions, which is
40
the leading cause of anthropogenic climate change. According to data from the United States
41
Energy Information Administration (EIA) online database, global carbon emissions increased
42
from 18.4 billion metric tonnes in 1980 to 32.7 billion metric tonnes in 2012, indicating an
43
increase of 77.5 percent over this period. Advanced countries such as the OECD members and
44
emerging economies such as Brazil, South Africa, China and India are responsible for such
45
phenomena increase. Thus, addressing the global challenge of climate change and global
46
warming requires drastic reduction of CO2 emissions.
47
One of the key measures aimed at addressing energy-related CO2 emission and mitigating climate
48
change and global warming is increasing the share of renewable energy (or reducing the share of
49
fossil-fuel energy) in the total energy mix. While considerable success has been achieved in this
50
respect in countries and regions like the United States, South Korea, Japan, Costa Rica and the
51
European Union, the level of renewable energy adoption in most developing countries is
52
relatively low. Net renewable electricity generation in the United States and Germany in 2011
53
stood at 527.49 TWh and 126.18 TWh respectively compared to 115.36 TWh for the entire
54
Africa continent. This corresponds to 13 percent, 22 percent and 18 percent of total electricity
55
generation in the United States, Germany and entire Africa respectively in the same period. More
56
interestingly, the development and adoption of renewable energy within developed and
57
developing countries varies considerably. For example, while Sweden had attained 65 percent of
58
its electricity consumption from renewable sources as at 2011, the United Kingdom had only
59
achieved 11 percent in the same period. Similarly, Brazil had achieved 86 percent of its
60
electricity requirement from renewable energy sources, mainly hydro, compared to 21 percent by
61
India as at the same period.
AC C
EP
TE D
M AN U
SC
RI PT
31
2
ACCEPTED MANUSCRIPT
The case of developing countries even requires more attention. IEA [1] states that the aggregate
63
energy-related CO2 emissions of non-Annex I developing countries surpassed those of Annex I
64
industrialised and transition countries (that have ratified the Kyoto Protocol) in 2008, and recent
65
growth in global carbon emissions is exclusively attributed to developing countries. It is therefore
66
expected that pro-poor growth in developing countries will result in a significant rise in energy
67
consumption and CO2 emissions in the near future (Gertler at al. [2]; Jakob et al [3]). Based on
68
these facts, there is need for a comprehensive analysis of the drivers and barriers to renewable
69
energy adoption in order to promote renewable energy use on a level that is needed to mitigate
70
climate change. This is in line with SDSN and IDDRI [4], which posits that deep decarbonisation
71
of the energy system requires both individual and globally coordinated decarbonisation efforts.
72
Numerous studies have been conducted on the econometric analysis of the drivers of and barriers
73
to renewable energy development and consumption in recent years (Brunnschweiler [5]; Dogan
74
[6, 7]; Omri and Nguyen [8]; Peterson [9]; Popp et al [10]; Sadorsky [11]; Salim et al [12]). The
75
main contribution of this paper to the literature is threefold. First, although there have been
76
enormous studies on renewable energy in the field of energy and environmental economics, the
77
majority of these studies focus specifically on developed and industrialised countries such as the
78
OECD countries and emerging countries such as China and India. In contrast, this paper analyses
79
the determinants of renewable energy adoption in a panel of both developed and developing
80
countries from various regions such as Africa, Europe, North and South America, Asia, taking
81
cognisance of differences between countries. The impact of economic and policy factors on
82
renewable energy adoption could differ between developed and developing countries, and may
83
also depend on specific development characteristics of each country.
84
Second, majority of previous papers in the literature study aggregate renewable energy,
85
combining hydro and non-hydro renewable energy. In our opinion, this is inappropriate for three
86
main reasons and may obscure the key factors influencing different types of renewable energy.
87
First, the factors that drive renewable energy from hydro and non-hydro are substantially
88
different. According to Johnstone et al [13], the effects of public policies on renewable energy
89
depend on the specific type of renewable energy. Hydroelectricity depends largely on the
90
availability of river flows while non-hydroelectricity such as solar, wind and geothermal does not
91
need this requirement. Given this stark difference in the technical and economic requirements for
92
hydro and non-hydroelectricity, conducting an econometric analysis of the factors driving
AC C
EP
TE D
M AN U
SC
RI PT
62
3
ACCEPTED MANUSCRIPT
aggregate/overall renewable energy may conceal the impact of economic and policy factors on
94
the specific type of renewable energy. Second, in a number of countries with substantial water
95
resources, hydro energy has already been highly exploited and contributes to electricity
96
generation, partly because it is a cheaper source of electricity production than fossil-based
97
generation. Countries like Brazil, Costa Rica and Nigeria generate a large proportion of their
98
electricity from hydro sources. It is preferable in these countries even without considering climate
99
change, unlike solar and wind renewables which are largely driven by climate change mitigation.
100
As an additional option for hydro energy-oriented countries as well as countries without
101
sufficient water resources, non-hydro renewable energy presents a key option. Third,
102
environmental concerns associated with hydro power plants could undermine the potential of this
103
energy source in the near future. It is recognised that hydro power plants cause disruptions in
104
rivers ecosystem and habitats. However, recent studies have shown that acute water shortages as
105
a result of global warming pose a serious challenge to hydro renewable energy in the future
106
(Schmidt et al [14]; van Vliet et al. [15]). Though non-hydro renewable energy sources such as
107
wind and biomass are also subject to the impacts of climate change, the effects are less compared
108
to hydro energy (Pasicko et al [16]). Hence, our focus on non-hydro renewable energy in this
109
paper. There is need to urgently develop non-hydro renewable energy in order to complement for
110
the potential shortfall that could emanate from the effects of global warming-induced water stress
111
on hydro energy contribution. We contribute to this strand of literature by disaggregating
112
renewable energy into hydro and non-hydro, and focusing our research on non-hydro renewable
113
energy sources.
114
The third contribution of this research to the literature is that it focuses on the share of non-hydro
115
renewable energy in total renewable energy, rather than the size of non-hydro renewable energy.
116
Increase in renewable energy consumption without commensurate reduction in fossil fuel energy
117
consumption will not achieve global climate objectives. Hence, we argue that as the consumption
118
of renewable energy increases, the share of fossil fuel consumption in total energy use should
119
reduce. Thus, our paper studies the SHARE (not the SIZE) of renewable energy in total energy
120
consumption. Most of the previous studies on renewable energy analyse the size/amount of
121
renewable energy produced or consumed (Brunnschweiler [5]; Omri and Nugyen [8]; Rafiq et al.
122
[17]; Salim and Rafiq [18]; Sardosky [11]). But it is the share (not the amount or size) of
123
renewable energy in total energy use that is important for climate change mitigation. Pfeiffer and
AC C
EP
TE D
M AN U
SC
RI PT
93
4
ACCEPTED MANUSCRIPT
Mulder [19] and Popp et al. [10], which are very similar to our study because they focus on non-
125
hydro renewable energy, use non-hydro renewable energy per capita and non-hydro (wind, solar
126
PV, geothermal and waste and biomass) net capacity respectively as the dependent variables. In
127
our opinion, using the share of renewables in total energy use as the indicator of energy transition
128
is more appropriate to determine progress towards decarbonisation of the energy system and
129
climate change mitigation. This is a strong departure from existing studies in the literature and a
130
key novelty of this paper. Therefore, this study centers on the share of non-hydro renewable
131
energy in total energy generation.
132
Based on this background, the objective of this paper is to empirically investigate the
133
determining factors of the share of non-hydro renewable energy using data of forty-six developed
134
and developing countries from different regions.
135
1.1
136
Studies on renewable energy have enjoyed distinguishable pedigree in the literature in recent
137
years. This is because adopting renewable energy sources is recognised as one of the major
138
means of decarbonising the energy system and mitigating climate change (Dogan and Seker [20];
139
Moomaw et al. [21]. Similar to this study, various works have been done to determine the factors
140
that influence renewable energy. Lin et al. [22] investigated the factors influencing renewable
141
energy consumption in China, focusing on the electricity sector. Using time series data from
142
1980 to 2011, Johansen cointegration method and error correction model, the study found that
143
financial and economic development significantly enhances renewable energy in China while
144
trade openness, FDI and lobby of fossil fuel undermine it. In another study focusing on
145
renewable energy in the United States, Sovacool [23] asserts that conventional energy sources
146
have potential impacts on the adoption of renewable energy, as utility firms are oriented towards
147
big conventional power plants. The drivers of renewable energy in European countries were
148
explored by Marques et al. [24]. The analysis employed fixed effect vector decomposition
149
(FEVD) technique on data covering the period 1990-2006; and focused on political,
150
socioeconomic and country-specific factors. Based on the results of the study, the goal of
151
reducing energy dependency encourages renewable energy while traditional energy sources and
152
CO2 emissions are not favourable to renewable energy. The determinants of renewable energy
153
consumption in six emerging economies (Turkey, Indonesia, India, Brazil, Philippines and
154
China) were studied by Salim and Rafiq [18]. The results of the study showed that income is a
SC
M AN U
AC C
EP
TE D
Brief review of the literature
RI PT
124
5
ACCEPTED MANUSCRIPT
key driver of renewables in all six countries. In addition, pollutant emissions significantly
156
influence renewable energy in Indonesia, Brazil, India and China. Sadorsky [11] analysed the
157
relationship between income and renewable energy in emerging economies, and found that
158
income has a significant impact on renewable energy consumption. In another study, Omri and
159
Nguyen [8] analysed the driving factors of renewable energy consumption in 64 countries. The
160
study divided the 64 countries into sub-panels of high, middle and low-income countries, and
161
employed the dynamic GMM estimation technique. Based on the results of the study, it was
162
shown that carbon emissions and trade openness are the major driving factors of renewable
163
energy. On the other hand, oil price has a small negative effect on renewable energy in the global
164
and middle-income panels.
165
In a survey study focusing on greenhouse gas mitigation technology transfer, Peterson [9] found
166
that there is little evidence on the impacts of FDI, ODA, trade and other funding sources.
167
Brunnschweiler [5] investigated the impact of financial sector development on renewable energy
168
in non-OECD countries, and the results showed a positive impact of commercial banking. In a
169
similar study, Glemarec [25] examine the role of finance in off-grid sustainable energy access for
170
the poor. Popp et al [10] assessed the factors that determine renewable energy development in 26
171
OECD countries using data from 1991 to 2004. Based on the analysis, they found that energy
172
security, fossil fuel production, future electricity demand and national renewable energy policies
173
do not have significant impact on renewable energy; while ratification of the Kyoto protocol and
174
deployment of low-carbon substitutes such as nuclear and hydro enhance renewable energy. The
175
relationship among renewable energy, GDP and energy prices in OECD countries was
176
investigated by Chang et al. [26]. They found that countries with higher GDP have the capacity
177
to adopt renewable energy despite their high prices. Rafiq et al. [17] analysed and compared the
178
relationship between renewable energy generation, income and CO2 emission in China and India
179
using multivariate vector error correction model.
180
Some studies have gone farther by analysing the factors that promote or hinder specific types of
181
renewable energy. The driving factors and effectiveness of policies to promote wind energy in
182
the United States are analysed by Bird et al. [27] and Menz and Vachon [28]. Huang et al. [29]
183
explore the barriers to wind energy. They find that apart from intermittency of wind energy, other
184
challenges include lack of national policies, poor institutional frameworks, and inadequate
185
economic assessments. In the analysis of non-hydro renewable energy in 108 countries, Pfeiffer
AC C
EP
TE D
M AN U
SC
RI PT
155
6
ACCEPTED MANUSCRIPT
and Mulder [19] found that income, education, democracy and implementation of economic and
187
regulatory instruments have significant positive effect on renewable energy diffusion. Garcia et
188
al. [30] use expert opinion analysis to investigate the potentials of renewable hydrogen storage
189
systems in Europe. Also studying the impact of policy risk on the development of solar PV in
190
Europe, Luthi and Wustenhagen [31] conduct a preference survey among European PV projects
191
developers and find that policy risks play an important role in investment decisions. The
192
summary of the current literature on the determinants of renewable energy use is shown in table
193
1.
AC C
EP
TE D
M AN U
SC
RI PT
186
7
ACCEPTED MANUSCRIPT
194
Table 1: Summary of current literature on the determinants of renewable energy use Scope/Coverage China
Method/Techniques Johansen’s cointegration and error correction model
Sovacool (2009)
United States
Interviews
Marques et al (2010)
European countries
Salim and Rafiq (2012)
Turkey, Indonesia, India, Brazil, China, Philippines 18 emerging countries 64 countries
Fixed effect vector decomposition (FEVD) technique Fully modified OLS (FMOLS), Dynamic OLS and Granger causality Panel cointegration and VECM Dynamic GMM
Peterson (2007)
Global
Literature survey
Brunnschweiler (2010)
119 Non-OECD countries
Panel data estimations
RE production, financial sector variables
Panel data analysis – pooled regression and fixed effects model
RE capacity, global knowledge stock, GDP per capita, growth rate of electricity consumption, Kyoto protocol, production of alternative energy sources
Popp et al (2011) 26 OECD countries
SC
M AN U
Share of RE in total energy use, political, socio-economic and country-specific factors RE consumption, real GDP, oil prices, CO2 emissions
TE D
RE consumption, income, electricity prices Amount of RE, CO2 emissions, oil price, trade openness, per capita GDP RE and variables used in other studies
EP
Omri and Nguyen (2014)
AC C
Sadorsky (2009)
Variables Used Share of renewable electricity in total electricity consumption, financial development, GDP, trade openness, FDI and fossil fuel “lobby effect” Wind energy
Findings Financial development and GDP enhances RE while trade openness, FDI and fossil fuel “lobby effect” undermine it
RI PT
Author(s) Lin et al (2016)
8
Socio-technical factors are responsible for neglect of RE in the US Goal of energy independence enhances RE while traditional energy sources and CO2 emissions impede RE Income have significant impact on RE in all the countries while pollutant emission affect RE in Brazil, China, India and Indonesia Increase in real income have significant positive effect on RE Trade openness and increase in CO2 emissions are the major drivers of RE Little evidence on the impacts of FDI, ODA, trade and other funding sources on greenhouse gas mitigation technology Adoption of Kyoto protocol and commercial banking has a positive influence on RE Ratification of the Kyoto protocol and deployment of nuclear and hydro enhances RE while energy security, fossil fuel production, future electricity demand and national RE policies do not have effect.
ACCEPTED MANUSCRIPT
OECD countries
Panel threshold regression model
Rafiq et al. (2014)
China and India
Multivariate VECM RE, GDP, CO2 emissions
Bird et al. (2005)
12 major wind power producing states in the US
Literature survey and case studies
Menz and Vachon (2006)
39 states in the US
Pooled OLS
Pfeiffer and Mulder (2013)
108 developing countries
Two-stage estimation method
Aguirre and Ibikunle (2014)
EU, OECD and BRICS countries
FEVD and Panel corrected standard errors (PCSE)
Source: Authors’ tabulation
M AN U
SC
Wind power, policy factors such as renewable portfolio standards (RPS), financial incentives, demand for green power, natural gas price volatility, wholesale market rules Wind power, state-level policies and retail choices
EP
TE D
Non-hydro renewable energy per capita, income, education, democracy, trade openness, aid, electricity consumption growth, FDI, energy mix, economic and regulatory instruments Share of RE in total energy use, political factors, socio-economic factors, country-specific factors
AC C
195
Contribution of RE to energy supply, GDP and CPI
9
Countries with higher GDP have the capacity to adopt RE despite their higher prices. The results are mixed for both India and China, and they differ in the short and long run Numerous drivers of wind RE influence one another’s effectiveness, but RPS is the most effective of state policy drivers.
RI PT
Chang et al. (2009)
Positive relationship between RPS and wind power and negative effect of retail choice Regulatory instruments, income, schooling and democracy promote RE while openness, aid, electrification and fossil fuel abundance delay the diffusion of RE.
CO2 emissions, high biomass and solar energy potential enhances RE; energy use, high electricity rates for the industrial sector, fossil fuel use have negative impacts on RE
ACCEPTED MANUSCRIPT
From the review of the current literature on the determinants and drivers of renewable energy,
197
some limitations are identified, which we aim to address in this paper. First, previous studies
198
focus extensively on aggregate renewable energy, with only few distinguishing between hydro
199
and non-hydro. We address this gap by focusing on non-hydro renewable energy. Second, most
200
of the papers use the aggregate amount/size of renewable energy in their analysis. We analyse the
201
driving factors of the share of non-hydro renewable energy in total energy use and argue that
202
economic and policy factors have different impacts on the size and share of renewable energy
203
consumption. Third, unlike previous studies that focus on developed and emerging economies,
204
the sample in this study consists of a panel of developed and developing countries from various
205
regions, subject to data availability. The inclusion of developed and developing countries in the
206
sample is to determine the universal factors that drive the share of non-hydro renewable energy.
207
The contents of the paper are as follows: Section one presents the background of the study and
208
summarises the relevant literature. The method of analysis is the focus of section two. Section
209
three presents the results of the empirical analysis while chapter four consists of the discussion of
210
results, policy implications and conclusion of the study.
211
2.0
Methodology
212
2.1
Variables, empirical model and data sources
213
The empirical analysis of this paper utilises panel data of forty-six developed and developing
214
countries over the period 1980-2011. The countries included in the study are listed in table 2
215
below. The variables included in this paper are based on theoretical understanding and from the
216
literature, subject to data availability. The dependent variable is the share of non-hydroelectricity
217
in total electricity generation. To confirm that there may be differences in the factors that
218
influence the share of non-hydro renewable electricity in total electricity generation and the
219
factors that influence the size of non-hydro electricity generation, we also use the latter as a
220
dependent variable in another model for comparison purpose.
221
Table 2: List of countries in the study Argentina Australia
Austria
Belize
Bolivia
Brazil
Canada
Chile
China
Colombia
Costa Rica
Denmark
Dominican Republic
Fiji
Finland
France
AC C
EP
TE D
M AN U
SC
RI PT
196
10
ACCEPTED MANUSCRIPT
Greece
Guatemala
India
Indonesia
Ireland
Italy
Jamaica
Japan
Jordan
Kenya
Mexico
Netherland
Nicaragua
Norway
New Zealand
Panama
Peru
Philippines
Singapore
Spain
Sweden
Thailand
Trinidad & Tobago
Turkey
Uruguay
United States of America
RI PT
Gabon
Senegal
Switzerland
United Kingdom
Source: Authors’ tabulation
223
To select the independent variables in this paper, we follow the general insights from the energy
224
economics literature. The independent variables in this study include foreign direct investment,
225
trade
226
commitment/institutions. Following Brunnschweiler [5], we include additional regressors - gross
227
fixed capital formation and natural resource dependence in the model - to check for the
228
robustness of the estimates of our variables of interest. Trade openness and foreign direct
229
investment are included in the model because trade linkages and foreign direct investment are
230
expected to enhance technology and knowledge transfer, and consequently facilitate the
231
development of renewable energy technologies (Omri and Nguyen [8]; Pfeiffer and Mulder [19]).
232
Trade openness is calculated as the share of import and export trade in GDP while foreign direct
233
investment is represented by the ratio of foreign direct investment to GDP. GDP per capita
234
represents level of development and capacity to invest in renewable energy technologies. It is
235
expected that as the level of income increases, citizens may begin to demand for more
236
environmental protection (Sardosky [11]). More importantly, improvement in the level of
237
development may enhance a country’s capability to invest in renewable energy and promote
238
sustainable development (Marques et al. [24]). This variable is measured by real GDP per capita.
239
Financial development is supposed to facilitate the development of renewable energy by
240
allocating financial resources to renewable energy investment and projects (Brunnschweiler [5]).
241
This is particularly because of the high investment costs of most renewable energy technologies.
242
Financial development in this study is indicated by domestic credit to the private sector as a
243
percent of GDP. Oil price indicates the price of substitutes. In the standard theory of demand and
per
M AN U
GDP
capita,
financial
development,
oil
price,
policy
AC C
EP
TE D
openness,
SC
222
11
ACCEPTED MANUSCRIPT
supply, the price of a substitute has impact on the demand or supply of a commodity. In this case,
245
we assume that conventional electricity sources from fossil fuel are substitutes for renewable
246
energy (Salim and Rafiq [18]). Thus, an increase in oil price will reduce the demand for oil and
247
fossil-fuel electricity generation and possibly increase the demand for renewable energy. Oil
248
price is the same for all countries but changes over time, so it is taken as a time-specific fixed
249
effect. So we hypothesise that an increase in oil price will lead to increase in renewable energy
250
generation. Oil price is proxied by West Texas Intermediate (WTI) spot prices. The Kyoto
251
protocol is used as a measure of institutional support and commitment to climate policy (Aguirre
252
and Ibikunle [32]; Pfeiffer and Mulder [19]). Like Aguirre and Ibikunle [32] and Popp et al. [10],
253
we adopt a dummy variable to capture it. The value of 1 represents ratification of the protocol
254
and 0 represents no ratification of the protocol. We expect a positive relationship between
255
ratification of the Kyoto protocol and renewable energy development.
256
Several literatures have shown that the more a country depends on fossil fuel, the less it is likely
257
to support the development of renewable energy. This is known as the “lobby effect” (Lin et al.
258
[22]; Marques at al. [24]). In past studies, the share of fossil fuel in total energy consumption is
259
used to capture the lobby effect. However, in this study we indicate it by the share of natural
260
resource/oil rent in GDP. This comprehensively captures the impact of natural resource/oil
261
dependence on renewable energy generation. The idea is to investigate if countries that depend
262
on natural resources, especially oil, see renewable energy adoption as a threat to the potentials of
263
their oil resources. Gross fixed capita formation indicates the level of existing physical
264
infrastructure and assets of a country. Countries with developed physical infrastructural assets
265
may be able to develop renewable energy better and faster than countries with lower physical
266
infrastructure. It is indicated as total gross fixed capital formation as a percentage of GDP. We
267
hypothesise a positive association between all the independent variables (except “lobby effect”)
268
and share of non-hydro renewable electricity generation. Table 3 below shows the variables, their
269
definitions and units of measurement.
270
AC C
EP
TE D
M AN U
SC
RI PT
244
Table 3: Definition of the variables used in the study Variable Definition SNHE Share of non-hydro power in total electricity generation NHE Amount/size of non-hydro power GDPPC GDP per capita 12
Units of measurement Percent TWh US$
ACCEPTED MANUSCRIPT
Percent Percent Percent Percent Percent US$ 0 or 1
RI PT
271
TO Trade openness (as % of GDP) FDI Foreign direct investment (as % of GDP) FIN Credit to the private sector (as % of GDP) GFCF Gross fixed capital formation (as % of GDP) RENT Share of resource/oil rents in GDP OIL WTI spot oil price KYOTO Ratification of the Kyoto Protocol (Dummy variable) Source: Authors’ tabulation
Thus, to investigate the dynamic relationship between the share of non-hydro renewable
273
electricity and its influencing factors, we assume a fixed effect specification following Kao et al.
274
[33] and specify the following baseline model:
SC
272
snheit = αgdppcit + δtoit + βfdiit + ρfinit + γoilt + ηkyotoit + εit………..1
276
To compare the impact of these variables on the share of non-hydro RE and the size of non-hydro
277
RE, we also estimate the following model:
M AN U
275
278
nheit = αgdppcit + δtoit + βfdiit + ρfinit + γoilt + ηkyotoit + εit………….2
279
To check for the robustness of the estimates, we follow Brunnschweiler [5] and include two
280
additional variables (gfcf and rent) to equation 1 and 2 to estimate the following model: snheit = αgdppcit + δtoit + βfdiit + ρfinit + γoilt + ηkyotoit + λgfcfit + φrentit + εit………..3.
282
Equation 1, 2 and 3 provide the mathematical expression of the model that describes the
283
relationship between non-hydro renewable energy and the selected variables. All the variables are
284
as described above in table 3. i refers to the forty-six countries included in the study; t refers to
285
the time period of the study (1980-2011); and εt is the error terms. The parameters in equation 1,
286
2 and 3 (α, δ, β, ρ, γ, η, λ, φ) are the coefficients of the corresponding variables. In other words,
287
they represent the effect of the corresponding variables on the dependent variable (nhe or shne).
288
For example, α, δ, β, ρ, γ, η, λ, φ are the respective estimated effects of GDP per capita, trade
289
openness, foreign direct investment, financial development, oil price, Kyoto protocol, gross fixed
290
capital formation and resource rent on the size or share of non-hydro renewable energy. If the
291
parameters are positive, it indicates that an increase in the value of the corresponding variables
292
has a positive effect on non-hydro renewable energy and vice-versa. The error term εt represents
293
other factors that affect non-hydro renewable energy but not included in the model.
294
The data used in the paper are obtained from various sources. SNHE is calculated based on
295
available data from the database of the United States Energy Information Administration (EIA).
296
Data on non-hydro renewable electricity and total electricity generation are obtained from this
AC C
EP
TE D
281
13
ACCEPTED MANUSCRIPT
database and the share of non-hydro renewable electricity is calculated. Data on GDP per capita,
298
trade openness, foreign direct investment, financial development, gross fixed capital formation
299
and natural resource rents are obtained from the World Development Indicator database of the
300
World Bank. Data on oil price is obtained from BP Statistical Review (BP [34]) while
301
information on the ratification, including date, of the Kyoto protocol is obtained from the United
302
Nations Framework Convention on Climate Change (UNFCCC [35]).
303
2.2
304
There are various techniques for estimating the coefficients of panel data models. The most
305
commonly used methods are the pooled OLS, fixed and random effect models estimators. But
306
these techniques are biased and produce inconsistent estimates when the variables are
307
cointegrated. This leads to the development of new estimators that estimate the cointegration
308
vectors of panel data. The new estimators include within and between-group estimators such as
309
OLS estimators, fully modified OLS (FMOLS) estimators and dynamic OLS (DOLS) estimators
310
(McCoskey and Kao [36]; Phillips and Moon [37]). They produce asymptotically unbiased,
311
normally distributed coefficient estimates (Pedroni [38]; Kao and Chiang [39]). The OLS and
312
DOLS are parametric methods, with the DOLS including the lagged first difference. On the other
313
hand, the FMOLS is a non-parametric method for correcting endogeneity (Chaiboonsri et al.
314
[40]). The OLS is biased and inconsistent when applied to cointegrated panels; but the FMOLS
315
and DOLS are more appropriate for heterogeneous cointegrated panels (Hamit-Haggar [41]). The
316
advantages of these estimation techniques are that they generates consistent estimates of the
317
parameters and also addresses the problems of endogeneity in the cointegration regression and
318
serial correlation in the residual. While Pedroni [38] find that DOLS has higher size distortions
319
than FMOLS, Kao and Chiang [39] show that FMOLS may be more biased than DOLS (Harris
320
and Sollis [42]). Ramirez [43] on the other hand argued that the FMOLS method is preferred
321
over the DOLS method for relatively small samples. In this study, we apply both the FMOLS and
322
DOLS techniques to determine the long run determinants of the share of non-hydro renewable
323
energy, following the procedure stated in box/stage 3 of the data flow diagram in figure 1.
324
We use the dynamic fixed effect estimator (DFE) to derive the short run dynamics of the
325
relationship among the variables, within the context of the error correction model. This technique
326
is based on recent developments in the literature on non-stationary cointegrated panel
RI PT
297
AC C
EP
TE D
M AN U
SC
Estimation technique
14
ACCEPTED MANUSCRIPT
econometrics (Blackburne and Frank [44]). It restricts the coefficient of the cointegrating vector,
328
the speed of adjustment coefficient and the short-run coefficient to be equal across the panels.
329
The error correction model distinguishes between the short and long run relationships among the
330
variables; and the coefficient of the error correction term indicates the speed of adjustment to
331
long run relationships. This procedure is implemented following box/stage 4 of the data flow
332
diagram. For more information on the FMOLS/DOLS, see Kao, Chiang and Chen [33]. See
333
Blackburne and Frank [44], Pesaran, Shin and Smith ([45], [46]) and see Pesaran and Smith [47]
334
for more information on the dynamic fixed effect (DFE).
335
Before carrying out the estimation, there is need to test for the stationarity of the variables as well
336
as the existence of a long run relationship between non-hydro renewable energy and its
337
influencing factors. According to Deng [48], most series of economic variables are non-
338
stationary and using such economic series for analysis would produce unreliable and biased
339
estimates. If the series contain unit roots, their first difference will be taken to eliminate the unit
340
root and make it stationary. Two categories of unit roots tests have been developed for panel
341
data. The first category is the tests that assume that the different cross-section sequences have a
342
common unit root process; and includes Levin-Lin-Chu (LLC) test (Levin et al [49]), Hadri LM
343
test (Hadri [50]) and Breitung test (Breitung [51]). On the other hand, the second category
344
assumes that the cross-section sequences have different individual unit root process; and they
345
include Im-Pesaran-Shin (IPS) test (Im et al [52]), Fisher-ADF and Fisher-PP tests (Choi [53]).
346
Because of their popularity and wide use in the literature, and given the heterogeneity among the
347
countries in this study, the second category tests are employed in this paper and the procedure is
348
shown in box/stage 1 of the data flow diagram. Furthermore, the cointegration test examines the
349
presence of a long run relationship among the variables. In this study, following the procedure in
350
box/stage 2 of the data flow diagram, we use three recently developed panel cointegration tests,
351
which includes Pedroni residual cointegration test (Pedroni [54]), Kao residual cointegration test
352
(Kao [55]) and Fisher-Johansen cointegration test (Maddala and Wu [56]) to examine the
353
existence of long run relationship between non-hydro renewable energy and its determinants.
354
The data flow diagram which presents an overview and steps of the methodological approach is
355
provided below in figure 1. We first conduct the panel unit root rests, followed by panel
356
cointegration test. After establishing a long run relationship, we use the FMOLS/DOLS to
AC C
EP
TE D
M AN U
SC
RI PT
327
15
ACCEPTED MANUSCRIPT
estimate the coefficient of the relationship and further use the DFE to estimate the short run
358
effect. The analysis is implemented using STATA and Eviews softwares. The data used for the
359
analysis in the various stages of the flow diagram is included as a supplementary information.
M AN U
SC
RI PT
357
360 361
Fig. 1: Data flow diagram that shows the steps and stages taken in the methodological approach.
362
3.0
363
The results of the analysis are presented and discussed in this section. This consists of the results
364
of the unit root tests, panel cointegration tests, pooled OLS tests, FMOLS and DOLS tests and
365
the robustness checks, following the sequence of the data flow diagram. This section further
366
discusses the result of the analysis and the findings.
367
3.1
368
The results of the panel unit root tests adopted are shown in table 4 below. Based on the result,
369
most of the variables are non-stationary at levels, but the first differences of all the variables are
370
stationary at 1% significance level. At levels, only FDI, GFCF and RENT are stationary because
371
the IPS, ADF and PP statistics for FDI, GFCF and RENT are significant at either 1, 5 or 10%
372
significance levels. At levels, there is existence of unit root for NHE, SHNE, GDPPC, TO, FIN,
373
OIL as the statistical values of the IPS, Fisher-ADF and PP tests are not significant (their
374
probability values are higher than 1, 5 and 10%). However, after taking the first difference of all
375
the variables, the statistics for all the variables are significant at 1% level. In other words, the
376
probability values of the statistics are less than 1%. This is consistent for the three unit root
EP
TE D
Results
AC C
Panel unit root tests
16
ACCEPTED MANUSCRIPT
377
testing methods used. The unit root test is not conducted for Kyoto because it is a dummy
378
variable that takes 0 and 1.
379
380 381
Table 4: Panel unit root tests Order of Variables IPS Fisher-ADF integration NHE 16.56088 9.6288 SNHE 12.0195 5.7474 GDPPC 8.6089 7.1174 TO 1.8871 2.3725 @levels FDI -2.7078*** -2.4227 FIN 3.1419 3.4297 OIL 17.4021 16.7592 GFCF -4.8136*** -4.8858*** RENT -6.0360*** -6.0410*** NHE -4.4904*** -3.6936*** SNHE -3.9407*** -5.7720*** GDPPC -7.6173*** -8.1801*** TO -13.9811*** -15.2547*** @first difference FDI -17.2587*** -18.6606*** FIN -10.2986*** -11.1284*** OIL -5.0961*** -5.20277*** GFCF -14.8877*** -16.2633*** RENT -12.2308*** -13.2823*** Source: Authors’ tabulation (results extracted from STATA software); Note: *** denotes 1% significance level.
382
3.2
383
The cointegration tests examine the existence of a long run relationships among the variables.
384
Given that all the variables are stationary after first differencing as shown in table 4, we test for
385
the existence of long run relationship among the variables using Pedroni, Kao and Johansen-
386
Fisher cointegration tests.
Fisher-PP
EP
AC C
387
Cointegration tests
TE D
M AN U
SC
RI PT
7.2648 1.7307 9.4946 1.6906 -10.5629*** 2.0277 15.3358 -6.5708*** -4.9371*** -19.5542*** -22.8941*** -16.2047*** -31.4283*** -42.0410*** -26.3115*** -41.0521*** -30.1946*** -37.1959***
Table 5: Pedroni cointegration test H0: No cointegration Panel v statistic Panel v weighted statistic Panel rho-statistic Panel rho weighted statistic Panel PP statistic Panel PP weighted statistic Panel ADF statistic Panel ADF weighted statistic Group rho statistic 17
Statistic -1.8883 -1.1408 2.2074 3.1393 -3.0500*** -1.6275* -3.4653*** -1.8760** 4.8264
p-value 0.97005 0.8730 0.9864 0.9992 0.0011 0.0518 0.0003 0.0303 1.0000
ACCEPTED MANUSCRIPT
388
Group PP statistic -2.0543** 0.0200 Group ADF statistic -2.6169*** 0.0044 Source: Authors’ tabulation (results extracted from Eviews software);
389
Note: *, **, *** denotes 10%, 5% and 1% significance level respectively The result of the Pedroni test in table 5 shows that the null hypothesis (H0) of no cointegration is
391
rejected for the panel PP, panel PP weighted, panel ADF, panel ADF weighted, group PP and
392
group ADF statistics. This is because the probability values associated with the statistics are less
393
than either 1, 5 or 10%. Out of the eleven statistics computed by the Pedroni test, six statistics
394
confirm the existence of panel cointegration among the variables as the statistics are significant
395
at either 1, 5 or 10% significance levels. According to Pedroni [57], the Panel ADF and Group
396
ADF are the most important statistics and should be the yardstick when the results are
397
conflicting.
400
SC
M AN U
399
Table 6: Kao cointegration test H0: No cointegration t-statistic p-value ADF -2.205965** 0.0137 Residual variance 2.806904 HAC variance 2.611304 Source: Authors’ tabulation (results extracted from Eviews software); Note: ** denotes 5% sig. level of the statistic
TE D
398
RI PT
390
The Kao residual and Johansen-Fisher panel cointegration tests are also conducted to confirm the
402
existence of cointegration in the variable. Based on the results of the Kao test (ADF statistics = -
403
2.2060, p = 0.0137) in table 6, the null hypothesis of no cointegration is rejected as the p-value
404
associated with the t-statistic is less than the conventional 5% significance level.
405
Table 7: Johansen Fisher panel cointegration test Trend No intercept and trend Intercept only scenarios Hypothesized Fisher stat Fisher stat Fisher stat Fisher stat no. of CE(s) (from trace (from max- (from trace (from maxstatistics) eigen statistics) eigen statistics) statistics) None 1504.0*** 936.7*** 1623.0*** 1099.0*** At most 1 789.2*** 397.7*** 856.7*** 456.6*** At most 2 464.1*** 232.6*** 468.8*** 256.9*** At most 3 296.2*** 161.2*** 264.2*** 161.1*** At most 4 199.3*** 135.4*** 158.6*** 110.1* At most 5 141.3*** 132.0*** 110.4* 98.19 At most 6 88.80 88.80 107.1 107.1
AC C
EP
401
18
Intercept and trend Fisher stat (from trace statistics) 2055.0*** 1177.0*** 622.0*** 353.5**8 216.9*** 149.3*** 98.58
Fisher stat (from maxeigen statistics) 1085.0*** 584.6*** 317.7*** 181.6*** 119.6** 115.9** 98.58
ACCEPTED MANUSCRIPT
Source: Authors’ tabulation (results extracted from Eviews software);
407
Note: *, **, *** denotes 10%, 5% and 1% sig. level respectively
408
The Johansen-Fisher cointegration test uses the trace statistics and maximum eigenvalues to test
409
the cointegration ranks among the panel series. The results of the Johansen-Fisher test in table 7
410
also show the existence of cointegration as the null hypothesis of no cointegration (none) is
411
rejected at 1% significance level under all the trend deterministic scenarios (no intercept and
412
trend, intercept only, and intercept and trend). The scenarios are included to check if there is
413
difference in the result if the model specification changes. The results of the three panel
414
cointegration tests are largely consistent and provide evidence of a long run relationship among
415
the variables.
416
3.3
417
The motivation for using the FMOLS/DOLS to obtain the long run estimates of the impacts of
418
the selected variables on the share of non-hydro renewable energy in this study is that they
419
address the problems of endogeneity and serial correlation in the model. Using conventional
420
pooled OLS to analyse the determinants of non-hydro renewable power produces unreliable
421
estimates. This is confirmed by the pooled OLS results in table 8 as it shows the presence of
422
serial correlation and heteroscedasticity. The Wooldridge test for autocorrelation (Drukker [58];
423
Wooldridge [59]) shows there is evidence of serial correlation as the null hypothesis of “no first
424
order autocorrelation” is rejected. Similarly, the null hypothesis of “constant variance” in the
425
Breusch-Pagan/Cook-Weisberg test is rejected. More so, there is reverse causality between the
426
dependent and independent variables in the model, which indicates the problem of endogeneity.
427
The results of the FMOLS and DOLS regressions are also presented in table 8. The results on the
428
left side of the table show the impacts of the variables on the share of non-hydro renewables in
429
total electricity generation (snhe) while the right hand side shows the impacts of the variables on
430
the size of non-hydro renewables (nhe). The results of both FMOLS and DOLS techniques,
431
particularly with respect to the sign, size and significance level, are very similar except that the
432
DOLS model has higher adjusted R-squared and lower standard errors and long run variance.
433
Table 8: FMOLS and DOLS regressions snhe – share of non-hydro RE Variables Pooled OLS
SC
RI PT
406
AC C
EP
TE D
M AN U
Long run effects: Fully Modified and Dynamic OLS
19
snhe – share of nonhydro RE
nhe – size of nonhydro RE
DOLS
DOLS
FMOLS
FMOLS
ACCEPTED MANUSCRIPT
-0.110 0.048*** 5.138*** -0.459 0.045*** 0.008 0.983 1.886 3.355
-0.192* 0.080*** 5.854** -2.880* 0.090*** -0.026 0.905 4.513 246.928
435
Note: *, **, *** denotes 10%, 5% and 1% sig. level respectively
436
From table 8, the results show that FDI has a negative but insignificant effect on non-hydro
437
renewable electricity generation. Specifically, FDI has a negative relationship with both the size
438
and share of non-hydro renewable energy, but the effect is insignificant in the long run. In other
439
words, FDI does not promote non-hydro renewable energy. On the contrary, the impact of
440
financial development on the size and share of non-hydro renewable electricity is positive and
441
significant, implying that financial development promotes non-hydro renewable energy. A 1%
442
increase in financial development leads to a 0.06% increase in the share of non-hydro renewable
443
energy and a 0.05-0.08% in the size of non-hydro renewable energy. GDP per capita, which
444
indicates the level of economic development, has a negative effect on the share of non-hydro
445
renewable electricity but has a positive impact on the amount of non-hydro renewable electricity
446
generation. A 1% increase in GDP per capita leads to a 5.14-5.85% increase in the size of non-
447
hydro renewable energy, but leads to a 3.22-2.87% decrease in the share of non-hydro renewable
448
energy. The impact of the Kyoto protocol on non-hydro renewable energy is inconsistent and
449
mostly insignificant. According to a priori expectation, oil price has a positive significant effect
450
on both the share and size of non-hydro renewable energy, indicating the substitutability between
451
fossil fuel and renewable energy. A 1% increase in oil price leads to a 0.05-0.09% increase in the
452
size of non-hydro renewable energy and a 0.02-0.03% increase in the share of non-hydro
453
renewable energy. Trade openness has a significant positive effect on the share of non-hydro
454
renewable electricity in total electricity generation but has an insignificant effect on the level of
AC C
EP
TE D
M AN U
SC
RI PT
434
FDI -0.060 0.031 -0.057 FIN 0.016*** 0.055*** 0.058*** GDPPC -1.269*** -3.222** -2.870*** KYOTO 0.846 1.395** 0.314 OIL 0.0037*** 0.030*** 0.021* TO 0.015*** 0.058*** 0.070*** 2 Adj. R 0.090 0.912 0.781 Standard error 6.792 2.112 3.333 Long run variance 4.639 39.978 Wooldridge serial corr. test 36.630*** Breusch-Godfrey/Cook375.64*** Weisberg Source: Authors’ tabulation (results extracted from Eviews software);
20
ACCEPTED MANUSCRIPT
455
non-hydro renewables. A 1% increase in trade openness leads to a 0.06-0.07% increase in the
456
share of non-hydro renewable energy.
457
The results of the robustness checks are presented in table 9. Two additional independent
458
variables – gfcf and rent - are added to the model to determine if the estimates of the initial
459
variables (fdi, fin, gdppc, kyoto, oil and to) changes significantly in terms of sign, size or level of
460
significance.
RI PT
3.4 Robustness checks
462
Table 9: Robustness checks of the FMOLS and DOLS estimates snhe – share of non-hydro nhe – size of non-hydro RE RE Variables DOLS FMOLS DOLS FMOLS FDI -0.043 -0.061 -0.052 -0.215* FIN 0.051*** 0.056*** 0.049*** 0.078*** GDPPC -2.039** -2.857*** 5.968*** 5.494** 0.679 -0.878 -1.266** -2.879* KYOTO OIL 0.026*** 0.024* 0.061*** 0.100*** TO 0.053*** 0.071*** 0.006 -0.025 GFCF -0.038 -0.017 -0.030 0.080 RENT -0.052 -0.0594 -0.180*** -0.147 Adj. R2 0.782 0.781 0.994 0.905 Standard error 3.326 3.328 1.154 4.510 Long run variance 31.655 39.851 0.512 244.885 Source: Authors’ tabulation (results extracted from Eviews software);
463
Note: *, **, *** denotes 10%, 5% and 1% sig. level respectively
TE D
M AN U
SC
461
Comparing the estimates of the initial variables in equation 1 and 2 (table 8) with that of equation
465
3 (table 9), we find that all the coefficient estimates of the variables do not change significantly
466
in terms of size, sign and level of significance. In other words, the inclusion of these additional
467
variables does not affect the estimation results of the baseline model, implying that the estimates
468
are not subject to model specification. The impact of gross fixed capital formation on the size
469
and share of non-hydro renewable electricity generation is not significant. This indicates that
470
there is no significant evidence that current level of physical infrastructural asset enhance the
471
development of non-hydro renewable electricity generation. Also, the evidence of the “lobby
472
effect” is not supported as the estimates are mostly insignificant. This contradicts Lin et al. [22]
473
and Marques et al. [24]. However, it is important to note that unlike in previous studies, we
474
investigate the impact of natural resource dependency on the desire for renewable energy. This
475
expands the idea of the fossil fuel industry “lobby” in the literature to the overall economic
AC C
EP
464
21
ACCEPTED MANUSCRIPT
interest of a country. The aim is to determine whether oil/resource dependent countries see
477
renewable energy as a threat to their economic sustenance. But there is no concrete evidence that
478
natural resource dependence impede the development of non-hydro renewable energy. Countries
479
that depend on natural resources, especially oil, for economic development do not necessarily see
480
renewable energy as a threat to the potential of their natural resources.
481
3.5
482
The result of the short run dynamics is presented in table 10 below.
Short-run dynamics: Panel error correction model
Table 10: Short run dynamics
483
RI PT
476
484
Std. z-statistic P>|z| Error C -0.468 2.189 -0.21 0.831 ∆FDI(-1) 0.009 0.014 0.68 0.499 ∆FIN(-1) -0.004 0.004 -1.23 0.218 ∆GDPPC(-1) -1.611 1.348 -1.19 0.232 ∆KYOTO(-1) -0.236 0.280 -0.84 0.400 ∆OIL(-1) -0.015*** 0.005 -3.04 0.002 ∆TO(-1) -0.001 0.006 -0.11 0.914 ECM(-1) -0.106*** 0.014 -7.70 0.000 Source: Authors’ tabulation (results extracted from Eviews software);
485
Note: *** denotes 1% sig. level
SC
Coefficient
TE D
M AN U
Variable
The result confirms the evidence of a long run relationship among the variables, as the ECM term
487
is significant and negative. The coefficient of the ECM (-0.106), which is the speed of
488
adjustment, implies that if there is a shock to the model, equilibrium will be restored at the speed
489
of 10.6% per year. Among all the variables, only oil price has a significant relationship with non-
490
hydro renewable energy in the short term.
491
4.0
492
The results of the study show that foreign direct investment has an insignificant effect on the
493
share and size of non-hydro renewable electricity generation both in the short and long run, as the
494
coefficients of foreign direct investment in tables 8 and 10 are not significant. This is contrary to
495
expectation but supports recent finding by Peterson [9] that there is little evidence that FDI
496
enhances renewable energy. Lin at al. [22] also finds that FDI undermine renewable energy in
497
China. Pfeiffer and Mulder [19] similarly show that increasing foreign direct investment
498
undermine the adoption and amount of non-hydro renewable energy. On the other hand, Kwakwa
499
[60] finds that foreign direct investment significantly impacts hydro power generation in Ghana.
AC C
Discussion
EP
486
22
ACCEPTED MANUSCRIPT
Foreign direct investment is expected to facilitate technology, skills and knowhow transfer,
501
which are expected to enhance local capacity to promote technology advancements; but there
502
seems to be no direct link between foreign direct investment and renewable energy generation.
503
On the contrary, foreign direct investment may increase overall energy consumption, possibly
504
from fossil fuel energy sources. Potential increase in overall energy consumption may crowd out
505
any increase in renewable energy, thereby undermining the effect on the share of renewable
506
energy in the energy mix. Furthermore, the pollution haven hypothesis postulates that foreign
507
direct investment may cause or aggravate environmental problems in their host countries. The
508
empirical evidence of a positive impact of foreign direct investment on the share of non-hydro
509
renewable electricity in total electricity generation is not strong. This could be also due to the
510
nature and composition of foreign direct investment, as different forms of foreign direct
511
investment may impact renewable energy in different ways.
512
Financial development has a strong positive impact on the amount and share of non-hydro
513
renewable electricity. The coefficient of financial development for both the share and size of
514
non-hydro renewable energy regardless of the estimation technique are positive and significant as
515
shown in table 8. This indicates that financial development contributes to both the level of non-
516
hydro renewable electricity and the share of non-hydro renewable electricity in total electricity
517
generation. This is in line with the findings of Brunnschweiler [5], which show that financial
518
intermediation, particularly commercial banking, has a significant positive effect on renewable
519
energy production, and the impact is large for non-hydropower renewable energy such as wind,
520
solar, geothermal and biomass. But this contrasts with the result of Pfeiffer and Mulder [19]
521
which find that access to finance has no significant effect on non-hydro renewable energy
522
diffusion. Given the high investment costs of renewable energy projects, access to finance is
523
essential to achieve high penetration and production of renewable energy. As the financial sector
524
develops, its capability to fund major long term projects and infrastructure, including renewable
525
energy projects, improves. But investing in renewable energy projects could be considered a risky
526
venture by the financial sector due to high initial cost outlays, long project payback periods as
527
well as risks associated with uncertainty about future global energy and climate change policies
528
(IEA [61]. Hence, the importance of the financial sector to the development of renewable energy
AC C
EP
TE D
M AN U
SC
RI PT
500
23
ACCEPTED MANUSCRIPT
should be recognised and necessary policies should be put in place to mitigate the risks
530
associated with the financing of renewable energy projects.
531
GDP per capita, which indicates the level of economic development, has a significant positive
532
effect on the size of non-hydro renewables and a significant negative effect on the share of non-
533
hydro renewable energy. According to table 8, the coefficient of GDP per capita for size of non-
534
hydro renewable energy is positive but is negative for the share of non-hydro renewable energy.
535
The positive effect of economic development on the size of non-hydro renewable energy in our
536
study corroborates the earlier results of Pfeiffer and Mulder [19]. But economic development
537
reduces the share of renewables in total electricity production. This implies that while economic
538
development may enhance the production of renewable energy, it may adversely affect the share
539
of renewables in total energy production. Economic development increases the capacity to invest
540
in renewable energy and other sustainable development policies and projects. However, it also
541
leads to increase in overall energy/electricity consumption (Saidi and Hammami [62]). This has
542
significant implications for policy. Economic development may increase the amount of non-
543
hydro renewable electricity generated, but in a lesser proportion than it increases overall
544
electricity generation. In other words, development leads to increase in energy use, but the
545
increase is met by increasing the supply of conventional fossil fuel electricity relative to non-
546
hydro renewable electricity. Economic development is expected to increase the capacity to invest
547
in renewable energy generation. However, it could also increase total electricity use relative to
548
the increase in non-hydro renewable electricity. Thus, economic development is not a sufficient
549
prerequisite to facilitate the transition to renewable energy.
550
Based on the coefficient of KYOTO in table 8, the impact of the Kyoto protocol on both the size
551
and share of non-hydro renewable energy is insignificant, implying that the ratification of the
552
Kyoto protocol has no impact on the acceleration of renewable energy. This is also in line with
553
the results of Pfeiffer and Mulder [19], which shows a very weak link between ratification of the
554
protocol and the probability of investing in non-hydro renewable energy. Almer and Winkler [63]
555
also find very little evidence for the effectiveness of the Kyoto Protocol. This might be because
556
the Kyoto protocol focuses less on renewable energy and technology policies in favour of other
557
market-oriented policies. To achieve the goals of the Kyoto protocol, three basic mechanisms -
558
joint implementation, clean development mechanism, and emissions trading - were provided for
AC C
EP
TE D
M AN U
SC
RI PT
529
24
ACCEPTED MANUSCRIPT
and applied by most industrialised countries. In the Kyoto protocol, there was no direct emphasis
560
on technology policies that would enhance the development of renewable energy technology. The
561
protocol’s trading mechanisms are inadequate to enhance renewable energy as argued by Driesen
562
[64], partly because they placed much emphasis on emission reduction through trading.
563
Oil price has a significant positive relationship with non-hydro renewable energy in the long run.
564
The coefficient of OIL in table 8 is positive and significant across methods. This shows that
565
increase in oil price has significant impact on improving both the amount of non-hydro
566
renewable energy production and the share of non-hydro renewables in total energy production.
567
This confirms that conventional fossil-fuel and renewable energy are substitutes in the long run
568
(Pfeiffer and Mulder [19]). Increase in the price of oil reduces the demand for it and enhances the
569
competitiveness and use of renewable energy. But, the impact of oil price increase on the level of
570
non-hydro power is greater than the impact on the share of non-hydro power, signifying that
571
increase in oil price increases the amount of non-hydro renewable power generation but not in
572
the same proportion that it reduces the amount of fossil fuel power generation. This indicates that
573
oil price increase boosts the production of renewable energy but reduces fossil fuel energy
574
demand in a lesser proportion. This underlines the dominant role that fossil fuel plays in the
575
current energy landscape and the possible difficulties that may be associated with phasing it out
576
in the near future.
577
Trade openness does not have significant impact on the size of non-hydro renewables (as the
578
coefficient in table 8 is not statistically significant) but has a significant positive impact on the
579
share of non-hydro renewable in total energy production. With respect to the insignificant effect
580
of trade openness on the size of non-hydro renewable energy, this study is in line with the
581
findings of Pfeiffer and Mulder [19]. On the contrary, Omri and Nguyen [8] find trade openness
582
to be an important driver of the amount of aggregate renewable energy. Trade openness is
583
expected to facilitate the exchange of green technologies and also creates opportunities for
584
human capital accumulation and development. This finding suggests that trade openness may not
585
show appreciable effect on the amount of non-hydro renewable electricity generation, but has a
586
significant impact on the share of non-hydroelectricity in total electricity generation in the long
587
term. This result is counterintuitive and should be taken with some caution. However, a plausible
588
explanation is that the impact of trade openness may depend on the composition and direction of
AC C
EP
TE D
M AN U
SC
RI PT
559
25
ACCEPTED MANUSCRIPT
trade as well as on the type of renewable energy. According to EIA [65], sectoral energy intensity
590
is heavily influenced by the composition of domestic and internationally traded goods and
591
services. In addition, the transmission mechanism of trade openness on renewable energy needs
592
to be comprehensively investigated.
593
From the results of the panel error correction model in table 10, all the variables except oil price
594
do not have a significant impact on non-hydro renewable energy in the short run. Contrary to
595
expectation, oil price has a significant negative relationship with non-hydro renewable energy in
596
the short run, indicating that oil price increase leads to reduction in the share of non-hydro
597
renewables. Marques et al. [24] also reached similar conclusion for EU member countries while
598
Omri and Nguyen [8] find the same results for middle-income countries and in a global panel
599
analysis. This could be as a result of speculation as increase in oil price could trigger oil
600
production relative to the consequent increase in the demand for oil substitutes. Also, in the
601
absence of climate change mitigation policies, increase in the price of oil may lead to the use of
602
alternative fossil fuels such as coal and natural gas rather than renewables, as argued by van
603
Ruijven and van Vuuren [66]. This is even more likely given that increasing the production of
604
renewable energy in response to oil price increase takes time compared to alternative fossil fuel
605
energy sources. The insignificance of most of the variables in the short run indicates that long
606
term planning is required for renewable energy transmission as policies may not show
607
appreciable impacts in the short term. This is obvious given the long time frame for the
608
development of renewable energy projects and infrastructure and the potential challenges
609
associated with phasing out of fossil fuel energy sources.
610
4.1
611
Based on the result of the study, some policy suggestions are provided.
612
•
EP
TE D
M AN U
SC
RI PT
589
AC C
Policy implications
According to the results in table 10, most of the factors, except oil price, have impacts only in
613
the long term. Hence, policy makers should take cognisance of the long term it takes for
614
policies to have significant impacts on renewable energy transition. Countries should develop
615
long term policy mechanisms and integrate renewable energy transmission with long term
616
economic development plans.
617 618
•
The results of the study in table 8 shows that economic factors have differing impacts on the size and share of renewable energy in total energy use. Therefore, policies should be focused
26
ACCEPTED MANUSCRIPT
619
on not only increasing the size of non-hydro renewable energy, but more importantly on
620
increasing the share of non-hydro renewable energy in total energy production. Thus, efforts
621
should be made to promote the factors that increase renewable energy consumption while
622
reducing conventional fossil fuel use. •
Comparing the impact of economic factors on the size and share of non-hydro renewable
RI PT
623
energy in this study with previous studies that analyse aggregate renewable energy, the study
625
shows that economic factors may have different impact on different types of renewable
626
energy. As a result, there should also be emphasis on customising policies for different forms
627
of renewable energy, rather than adopting a one-size-fit-all policy. In other words, policy
628
makers should recognise the potential different impacts of economic factors and policy
629
instruments on different forms of renewable energy and direct targeted policies for the
630
prioritised form of renewable energy. •
M AN U
631
SC
624
Financial development is a key factor promoting both the size and share of non-hydro
632
renewable energy. Hence, development of the financial sector to enhance access to finance
633
for renewable energy investors should be a key priority and policy action for decision makers.
634
•
Increase in the price of oil also increases the share of non-hydro renewable energy in the long term, indicating that renewable energy may serve as a substitute for oil (or fossil fuel) if there
636
is a sustained increase in oil price making renewable energy price to be competitive. Hence,
637
to promote renewable energy, policy makers should eliminate artificial mechanisms like
638
fossil fuel subsidies that keeps oil price down (and by implications reduce the price of fossil-
639
fuel electricity source). •
EP
640
TE D
635
It is also important to note that most of economic and policy factors have long run relationship and affect one another. This is confirmed by the existence of a cointegration
642
relationship in table 5, 6 and 7. This should be taken into consideration when planning and
643
designing policies to facilitate renewable energy transition.
AC C
641
644
4.2
Conclusion
645
This paper investigates the driving factors of non-hydro renewable energy in a panel of forty-six
646
developed and developing countries from various regions, using panel cointegration techniques
647
and error correction model. The results show that trade openness, financial development, foreign
648
direct investment, economic development and ratification of the Kyoto protocol do not have
27
ACCEPTED MANUSCRIPT
impact on non-hydro renewable energy in the short run. In the long run, trade openness enhances
650
the share of non-hydro renewables in total electricity generation. Increase in oil price and
651
financial development have the effect of increasing both the amount and share of non-hydro
652
renewable electricity in total electricity generation. Economic development has a positive
653
significant impact on enhancing the amount of non-hydro renewables but has a negative effect on
654
the share of non-hydro renewables in total electricity generation. This is because economic
655
development may not increase renewable energy in the same or higher proportion than increase
656
in overall energy consumption. On the contrary, ratification of the Kyoto protocol and foreign
657
direct investment does not have significant effect on non-hydro renewable energy in the long run.
658
In summary, the results of this study indicate that economic variables and policies may affect
659
types of renewable energy differently as well as their size and share in total energy use.
660
This paper examines the driving forces of renewable energy transition, focusing on non-hydro
661
renewable energy. Future studies should further examine and possibly compare the impact of
662
these driving forces on both hydro and non-hydro renewable energy. In addition, the impact of
663
the variables could also differ between different sub-types of non-hydro renewable energy such
664
as solar, wind, wave, geothermal, etc. Thus, future research could also analyse the possibility.
665
Future studies also need to investigate the composition and structure as well as the transmission
666
mechanism of the impact of trade openness and foreign direct investment on renewable energy.
667
TE D
M AN U
SC
RI PT
649
Acknowledgments
669
The paper emanates from the co-author’s (O. E. Omoju) research visit to the Institute for
670
Sustainable Economic Development of the University of Natural Resources and Life Sciences
671
(BOKU), Vienna, Austria. The research visit was supported by the Ernst Mach Grant of the
672
Austrian Agency for International Cooperation in Education and Research (OeAD-GmbH). The
673
authors particularly acknowledge the scientific contributions and review comments of the hosts -
674
Prof. Klaus Salhofer, Dr. Ulrich Morawetz and Dr. Johannes Schmidt - and other participants at
675
the Institute’s seminars. We also thank the two anonymous reviewers for their comments and
676
suggestions on the paper.
AC C
EP
668
677 678
28
ACCEPTED MANUSCRIPT
679 680 681
References
683
[1] IEA. CO2 emissions from fuel combustion – Highlights. International Energy Agency, Paris,
684
2010.
685
[2] P. Gertler, O. Shelef, C. Wolfram, A Fuchs, How pro-poor growth affects the demand for
686
energy. NBER Working Paper Series No. 19092, National Bureau of Economic Research,
687
Cambridge, MA, 2013.
688
[3] M. Jakob, M. Haller, R. Marchinski, Will history repeat itself? Economic convergence and
689
convergence in energy use patterns, Energy Economics 34(2011) 95-104.
690
[4] SDSN, IDDRI, Pathways to deep decarbonisation - 2014 interim report. Sustainable
691
Development Solutions Network (SDSN) and Institute for Sustainable Development and
692
International Relations (IDDRI), Columbia University, New York, 2014.
693
[5] C.N. Brunnschweiler, Finance for renewable energy: An empirical analysis of developing and
694
transition economies, Environment and Development Economics 15(2010) 241-274.
695
[6] E. Dogan, Economic growth and electricity consumption from renewable and non-renewable
696
sources: A study of Turkey, Renewable & Sustainable Energy Reviews 52(2015) 534-546.
697
[7] E. Dogan, Analyzing the linkage between renewable and non-renewable energy consumption
698
and economic growth by considering structural break in time-series data, Renewable Energy
699
99(2016) 1126-1136.
700
[8] A. Omri, D.K. Nguyen, On the determinants of renewable energy consumption: International
701
evidence. Energy 72(2014) 554-560.
702
[9] S. Peterson, Greenhouse gas mitigation in developing countries through technology transfer?
703
A survey of empirical evidence, Mitigation and Adaptation Strategies for Global Change
704
13(2007) 283-305.
705
[10] D. Popp, I. Hascic, N. Medhi, Technology and the diffusion of renewable energy, Energy
706
Economics 33(2011) 648-662.
707
[11] P. Sardosky, Renewable energy consumption and income in emerging economies, Energy
708
Policy 37(2009) 4021-4028.
AC C
EP
TE D
M AN U
SC
RI PT
682
29
ACCEPTED MANUSCRIPT
[12] R.A. Salim, K. Hassan, S. Shafiei, Renewable and non-renewable energy consumption and
710
economic activities: Further evidence from OECD countries, Energy Economics 44(2014) 350-
711
360.
712
[13] N. Johnstone, I. Hascic, D. Popp, Renewable energy policies and technological innovation:
713
Evidence based on patent counts, Environmental and Resource Economics 45(2010) 133-155.
714
[14] J. Schmidt, R. Cancella, A.O. Pereira Jnr, The effect of windpower on long-term variability
715
of combined hydro-wind resources: The case of Brazil. Renewable and Sustainable Energy
716
Reviews 55(2016) 131-141.
717
[15] M.T.H. van Vliet, D. Wiberg, S. Leduc, K. Riahi, Power-generation system vulnerability
718
and adaptation to changes in climate and water resources, Nature Climate Change 6(2016) 375-
719
380.
720
[16] R. Pasicko, C. Brakovic, Z. Simic, Assessment of climate change impacts on energy
721
generation from renewable sources in Croatia, Renewable Energy 46(2012) 224-231.
722
[17] S. Rafiq, H. Bloch, R. Salim, Determinants of renewable energy adoption in China and
723
India: A comparative analysis, Applied Economics 46(2014) 2700-2710.
724
[18] R.A. Salim, S. Rafiq, Why do some emerging economies proactively accelerate the adoption
725
of renewable energy, Energy Economics 34(2012) 1051-1057.
726
[19] B. Pfiffer, P. Mulder, Explaining the diffusion of renewable energy technology in
727
developing countries, Energy Economics 40(2013), 285-296.
728
[20] E. Dogan, F. Seker, The influence of real output, renewable and non-renewable energy,
729
trade and financial development on carbon emissions in the top renewable, Renewable &
730
Sustainable Energy Reviews 60(2016) 1074-1085.
731
[21] W. Moomaw, F. Yamba, M. Kamimoto, L. Maurice, J. Nyboer, K. Urama, T. Weir,
732
Introduction, in: O. Edenhofer, R. Pichs-Madruga, Y. Sokona, K. Seyboth, P. Matschoss, S.
733
Kadner, T. Zwickel, P. Eickemeier, G. Hansen, S. Schlomer, C. von Stechow (Eds.), IPCC
734
Special Report on Renewable Energy Sources and Climate Change Mitigation, Cambridge
735
University Press, New York, 2011.
736
[22] B. Lin, O.E. Omoju, U.J. Okonkwo, Factors influencing renewable electricity consumption
737
in China, Renewable & Sustainable Energy Reviews 55(2016) 687-696.
AC C
EP
TE D
M AN U
SC
RI PT
709
30
ACCEPTED MANUSCRIPT
[23] B. Sovacool, Rejecting renewables: The socio-technical impediments to renewable
739
electricity in the United States, Energy Policy 37(2009) 4500-4513.
740
[24] A.C. Marques, J.A. Fuinhas, J.R.P. Manso, Motivations driving renewable energy in
741
European countries: A panel data approach, Energy Policy 38(2010) 6877-6885.
742
[25] Y. Glemarec, Financing off-grid sustainable energy access for the poor. Energy Policy
743
47(2012) 87-93.
744
[26] T.H. Chang, C.M. Huang, M.C. Lee, Threshold effect of the economic growth rate on the
745
renewable energy development from a change in energy price: Evidence from OECD countries,
746
Energy Policy 37(2009) 5796-5802.
747
[27] L. Bird, M. Bolinger, T. Gagliano, R. Wiser, M. Brown, B. Parsons, Policies and market
748
factors driving wind power development in the United States, Energy Policy 33(2005) 1397-
749
1407.
750
[28] F. Menz, S. Vachon, The effectiveness of different policy regimes for promoting wind
751
power: Experience from the States, Energy Policy 34(2006) 1786-1796.
752
[29] S.C. Huang, S.L. Lo, Y.C. Lin, To re-explore the causality between barriers to renewable
753
energy development: A case study of wind energy, Energies 6(2013) 4465-4488.
754
[30] D.A. Gracia, F. Barbanera, F. Cumo, U. Di Matteo, B. Nastasi, Expert opinion analysis on
755
renewable hydrogen storage systems potential in Europe, Energies 9(2016) 1-22.
756
[31] S. Luthi, R. Wustenhagen, The price of policy risk – Empirical insights from choice
757
experiments with European photovoltaic project developers, Energy Economics 34(2012) 1001-
758
1011.
759
[32] M. Aguirre, G. Ibikunle, Determinants of renewable energy growth: A global sample
760
analysis, Energy Policy, 69(2014) 374-384.
761
[33] C. Kao, M.H. Chiang, B. Chen, International R&D spillovers: An application of estimation
762
and inference in panel cointegration, Oxford Bulletin of Economics and Statistics 61(1999) 693-
763
711.
764
[34] BP, BP statistical review of world energy, BP Plc, London, 2016.
765
[35] UNFCCC, Kyoto Protocol - Status of ratification, United Nations Framework Convention
766
on
AC C
EP
TE D
M AN U
SC
RI PT
738
Climate
Change,
31
Bonn,
2009.
ACCEPTED MANUSCRIPT
http://unfccc.int/files/kyoto_protocol/status_of_ratification/application/pdf/kp_ratification.pdf
768
(accessed 23.06.16)
769
[36] S. McCoskey, C. Kao, Comparing panel data cointegration tests with an application of the
770
twin deficits problem, Mimeo, Syracuse University, New York, 1999.
771
[37] P.C.B. Phillips, H. Moon, Linear regression limit theory for non-stationary panel data,
772
Econometrica 67(1999) 1057–1111.
773
[38] P. Pedroni, Fully modified OLS for heterogeneous cointegrated panels: in B.H. Baltagi, T.B.
774
Fomby, R.C. Hill (Eds.), Nonstationary Panels, Panel Cointegration, and Dynamic Panels
775
(Advances in Econometrics, Volume 15), Elsevier Science Inc., 2000, pp. 93-130.
776
[39] C. Kao, M.H. Chiang, On the estimation and inference of a cointegrated regression in panel
777
data, Advances in Econometrics 15(2000) 179-222.
778
[40] C. Chaiboonsri, J. Sriboonjit, T. Sriwichailamphan, P. Chaitip, S. Sriboonchitta, A panel
779
cointegration analysis: An application to international tourism demand of Thailand, Annals of the
780
University of Petrosani, Economics, 10(2010), 69-86.
781
[41] M. Hamit-Haggar, Greenhouse gas emissions, energy consumption and economic growth: A
782
panel cointegration analysis from Canadian industrial sector perspective, Energy Economics
783
34(2012) 358-364.
784
[42] R. Harris, R. Sollis, Applied time series modeling and forecasting. John Wiley & Sons Ltd,
785
England, 2005.
786
[43] M.D. Ramirez, A panel unit root and panel cointegration test of the complementarity
787
hypothesis in the Mexican case: 1960-2001, Atlantic Economic Journal 35(2007) 343-356.
788
[44] E.F. Blackburne, M.W. Frank, Estimation of nonstationary heterogeneous panels, The
789
STATA Journal 7(2007) 197-208.
790
[45] M.H. Pesaran, Y. Shin, R.P. Smith, Estimating long-run relationships in dynamic
791
heterogeneous panels, DAE Working Papers Amalgamated Series No. 9721, 1997.
792
[46] M.H. Pesaran, Y. Shin, R.P. Smith, Pooled mean group estimation of dynamic
793
heterogeneous panels, Journal of the American Statistical Association 94(1999) 621-634.
794
[47] M.H. Pesaran, R.P. Smith, Estimating long-run relationships from dynamic heterogeneous
795
panels, Journal of Econometrics 68(1995) 79-113.
AC C
EP
TE D
M AN U
SC
RI PT
767
32
ACCEPTED MANUSCRIPT
[48] A. Deng, Understanding spurious regression in financial economics, Journal of Financial
797
Econometrics, 12(2014) 122-150.
798
[49] A. Levin, C.F. Lin, C.S.J Chu, Unit root tests in panel data: Asymptotic and finite-sample
799
properties, Journal of Econometrics 108(2002) 1-24.
800
[50] K. Hadri, Testing for stationarity in heterogeneous panel data, Econometrics Journal 3(2000)
801
148-161.
802
[51] J. Breitung, The local power of some unit root tests for panel data, in: B.H. Baltagi (ed.),
803
Nonstationary Panels, Panel Cointegration, and Dynamic Panels: Advances in Econometrics,
804
Volume 15: JAY Press, Amsterdam, 2000, pp. 161-178.
805
[52] K.S. Im, M.H. Pesaran, Y. Shin, Testing for unit roots in heterogeneous panels, Journal of
806
Econometrics 115(2003) 53-74.
807
[53] I. Choi, Unit root tests for panel data, Journal of International Money and Finance 20(2001)
808
249-272.
809
[54] P. Pedroni, Panel cointegration: Asymptotic and finite sample properties of pooled time
810
series tests with an application to the PPP hypothesis, Econometric Theory 20(2004) 597-625.
811
[55] C. Kao, Spurious regression and residual-based tests for cointegration in panel data when
812
the cross-section and time series dimensions are comparable, Journal of Econometrics 90(2004)
813
1-44.
814
[56] G.S Maddala, S. Wu, A comparative study of unit root tests with panel data and a new
815
simple test, Oxford Bulletin of Economics and statistics 61(1999) 631-652.
816
[57] P. Pedroni, Critical values for cointegration tests in heterogeneous panels with multiple
817
regressors, Oxford Bulletin of Economics and Statistics 61(1999), 653-670.
818
[58] D.M. Drukker, Testing for serial correlation in linear panel-data models, Stata Journal
819
3(2003) 1-10.
820
[59] J.M. Wooldridge, Econometric analysis of cross section and panel data. The MIT Press,
821
Cambridge, Massachusetts, 2002.
822
[60] P.A. Kwakwa, An investigation into the determinants of hydropower generation in Ghana,
823
MPRA Working Paper No. 68033, Munich Personal RePEc Archive, Munich, 2015.
824
[61] IEA, 2005. Reducing greenhouse gas emissions: The potential of coal. Coal Industry
825
Advisory Board, International Energy Agency, Paris.
AC C
EP
TE D
M AN U
SC
RI PT
796
33
ACCEPTED MANUSCRIPT
[62] K. Saidi, S. Hammami, The impact of CO2 emissions and economic growth on energy
827
consumption in 58 countries, Energy Reports 1(2015) 62-70.
828
[63] C. Almer, R. Winkler, Analysing the effectiveness of international environmental policies:
829
The case of the Kyoto Protocol, Journal of Environmental Economics and Management 82(2017)
830
125-151.
831
[64] D.M. Driesen, Renewable energy under the Kyoto protocol: The case for mixing
832
instruments. College of Law Faculty Scholarship Paper No. 27, Syracuse University, 2007.
833
[65] EIA, Annual energy outlook 2016 with projections to 2040, Energy Information
834
Administration, Washington, D.C., 2016.
835
[66] B. van Ruijven, D.P. van Vuuren, Oil and natural gas prices and greenhouse gas emission
836
mitigation, Energy Policy 37(2009) 4797-4808.
AC C
EP
TE D
M AN U
SC
RI PT
826
34
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
837
35
ACCEPTED MANUSCRIPT
EP
TE D
M AN U
SC
RI PT
We examine the driving factors of non-hydro renewable energy The study focuses on both the size and share of non-hydro renewable energy Oil price and financial development have impact on non-hydro renewable energy Economic factors impact the size and share of non-hydro renewables in different ways The impacts of economic and policy factors are mostly in the long run
AC C
• • • • •